U.S. patent application number 16/537542 was filed with the patent office on 2019-11-28 for systems and methods for full body measurements extraction using multiple deep learning networks for body feature measurements.
The applicant listed for this patent is Bodygram, Inc.. Invention is credited to Kyohei Kamiyama, Chong Jin Koh.
Application Number | 20190357615 16/537542 |
Document ID | / |
Family ID | 66825815 |
Filed Date | 2019-11-28 |
![](/patent/app/20190357615/US20190357615A1-20191128-D00000.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00001.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00002.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00003.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00004.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00005.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00006.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00007.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00008.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00009.png)
![](/patent/app/20190357615/US20190357615A1-20191128-D00010.png)
View All Diagrams
United States Patent
Application |
20190357615 |
Kind Code |
A1 |
Koh; Chong Jin ; et
al. |
November 28, 2019 |
SYSTEMS AND METHODS FOR FULL BODY MEASUREMENTS EXTRACTION USING
MULTIPLE DEEP LEARNING NETWORKS FOR BODY FEATURE MEASUREMENTS
Abstract
Disclosed are systems and methods for full body measurements
extraction using a mobile device camera. The method includes the
steps of receiving one or more user parameters from a user device;
receiving at least one image from the user device, the at least one
image containing the human and a background; performing body
segmentation on the at least one image to identify one or more body
features associated with the human from the background; performing
annotation on the one or more identified body features to generate
annotation lines on each body feature corresponding to body feature
measurement locations utilizing a plurality of annotation
deep-learning networks that have been separately trained on each
body feature; generating body feature measurements from the one or
more annotated body features utilizing a sizing machine-learning
module based on the annotated body features and the one or more
user parameters; and generating body size measurements by
aggregating the body feature measurements for each body
feature.
Inventors: |
Koh; Chong Jin; (Las Vegas,
NV) ; Kamiyama; Kyohei; (Tokyo, JP) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Bodygram, Inc. |
Las Vegas |
NV |
US |
|
|
Family ID: |
66825815 |
Appl. No.: |
16/537542 |
Filed: |
August 10, 2019 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
PCT/US19/27564 |
Apr 15, 2019 |
|
|
|
16537542 |
|
|
|
|
16195802 |
Nov 19, 2018 |
10321728 |
|
|
PCT/US19/27564 |
|
|
|
|
62809218 |
Feb 22, 2019 |
|
|
|
62660377 |
Apr 20, 2018 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06N 7/005 20130101;
G06N 20/10 20190101; G06T 7/73 20170101; G06T 17/10 20130101; G06T
2207/20104 20130101; G06N 20/00 20190101; G06N 3/0454 20130101;
G06N 3/126 20130101; G06T 2207/20081 20130101; G06T 7/11 20170101;
G06T 2207/10004 20130101; G06T 11/00 20130101; G06N 3/08 20130101;
G06T 2207/30196 20130101; G06N 5/003 20130101; G06T 7/62 20170101;
G06N 5/048 20130101; G06N 20/20 20190101; G06T 2200/24 20130101;
G06T 11/203 20130101; G06T 2207/20084 20130101; G06T 7/143
20170101; G06K 9/00369 20130101; A41H 1/02 20130101 |
International
Class: |
A41H 1/02 20060101
A41H001/02; G06T 11/20 20060101 G06T011/20; G06N 20/00 20060101
G06N020/00; G06T 7/143 20060101 G06T007/143; G06T 17/10 20060101
G06T017/10; G06K 9/00 20060101 G06K009/00 |
Claims
1. A computer-implemented method for generating body size
measurements of a human, the computer-implemented method executable
by a hardware processor, the method comprising: receiving one or
more user parameters from a user device; receiving at least one
image from the user device, the at least one image containing the
human and a background; performing body segmentation on the at
least one image to identify one or more body features associated
with the human from the background; performing annotation on the
one or more identified body features to generate annotation lines
on each body feature corresponding to body feature measurement
locations utilizing a plurality of annotation deep-learning
networks that have been separately trained on each body feature;
generating body feature measurements from the one or more annotated
body features utilizing a sizing machine-learning module based on
the annotated body features and the one or more user parameters;
and generating body size measurements by aggregating the body
feature measurements for each body feature.
2. The method of claim 1, wherein the performing body segmentation
step comprises: generating a segmentation map of the body features
on the human; and cropping the one or more identified body features
from the human and the background based on the segmentation map,
wherein the cropped body features are used in the performing
annotation step.
3. The method of claim 2, wherein the performing body segmentation
step utilizes a segmentation deep-learning network that has been
trained on segmentation training data, wherein the segmentation
training data comprise one or more images for one or more sample
humans and a body feature segmentation for each body feature for
the one or more sample humans.
4. The method of claim 3, wherein the body feature segmentation is
extracted underneath clothing, and wherein the segmentation
training data comprises body segmentation estimating the sample
human's body underneath the clothing.
5. The method of claim 1, wherein the annotation deep-learning
networks utilize annotation training data comprising one or more
sample images for one or more sample humans and an annotation line
for each body feature for the one or more sample humans.
6. The method of claim 5, wherein the annotation line on each body
feature comprises one or more line segments corresponding to a
given body feature measurement, and wherein the generating body
feature measurements step utilizes the annotation line on each body
feature.
7. The method of claim 6, wherein the at least one image comprises
at least a front-view image and a side-view image of the human, and
wherein the method further comprises the following steps after the
performing annotation step: calculating at least one circumference
of at least one annotated body feature utilizing line-annotated
front-view and side-view images and a height of the human; and
generating the body feature measurements from the at least one
circumference utilizing the sizing machine-learning module based on
the at least one circumference, the height, and the one or more
user parameters.
8. The method of claim 1, wherein the sizing machine-learning
module comprises a random forest algorithm, and wherein the sizing
machine-learning module is trained on ground truth data comprising
one or more sample body feature measurements for one or more sample
humans.
9. The method of claim 1, wherein the user parameters are selected
from the group consisting of a height, a weight, a gender, an age,
and a demographic information associated with the human.
10. The method of claim 9, wherein the receiving the one or more
user parameters from the user device comprises receiving user input
of the user parameters through the user device.
11. The method of claim 9, wherein the receiving the one or more
user parameters from the user device comprises receiving a
measurement performed by the user device.
12. The method of claim 1, wherein the at least one image is
selected from the group consisting of a front-view image of the
human and a side-view image of the human.
13. The method of claim 12, wherein the at least one image further
comprises an additional image of the human taken at an angle of
approximately 45 degrees with respect to the front-view image of
the human.
14. The method of claim 1, wherein the performing body segmentation
step on the at least one image further comprises receiving user
input to increase an accuracy of the body segmentation, and wherein
the user input comprises a user selection of one or more portions
of the identified body features that correspond to a given region
of the human's body.
15. The method of claim 1, wherein the at least one image comprises
at least one image of a fully-clothed user or a partially-clothed
user, and wherein the generating body feature measurements step
further comprises generating the body feature measurements on the
at least one image of the fully-clothed user or the
partially-clothed user.
16. The method of claim 1, wherein the body size measurements
comprise first body size measurements, wherein the method further
comprises generating second body size measurements using a second
machine-learning module, and wherein an accuracy of the second body
size measurements is greater than an accuracy of the first body
size measurements.
17. The method of claim 1, further comprising: determining whether
a given body feature measurement of the identified body features
corresponds to a confidence level below a predetermined value; and
in response to determining that the given body feature measurement
corresponds to a confidence level below the predetermined value,
performing 3D model matching using a 3D model matching module on
the identified body features to determine a matching 3D model of
the human, wherein one or more high-confidence body feature
measurements are used to guide the 3D model matching module,
performing body feature measurements based on the matching 3D
model, and replacing the given body feature measurement with a
projected body feature measurement from the matching 3D model.
18. The method of claim 1, further comprising: determining whether
a given body feature measurement of the identified body features
corresponds to a confidence level below a predetermined value; and
in response to determining that the given body feature measurement
corresponds to a confidence level below the predetermined value,
performing skeleton detection using a skeleton detection module on
the identified body features to determine joint positions of the
human, wherein one or more high-confidence body feature
measurements are used to guide the skeleton detection module,
performing body feature measurement based on the determined joint
positions, and replacing the given body feature measurement with a
projected body feature measurement from the skeleton detection
module.
19. The method of claim 1, further comprising: pre-processing the
at least one image of the human and the background before the
performing body segmentation step, wherein the pre-processing
comprises at least a perspective correction on the at least one
image, and wherein the perspective correction is selected from the
group consisting of perspective correction utilizing a head of the
human, perspective correction utilizing a gyroscope of the user
device, and a perspective correction utilizing another sensor of
the user device.
20. A computer program product for generating body size
measurements of a human, comprising a non-transitory computer
readable storage medium having program instructions embodied
therein, the program instructions executable by a processor to
cause the processor to: receive one or more user parameters from a
user device; receive at least one image from the user device, the
at least one image containing the human and a background; perform
body segmentation on the at least one image to identify one or more
body features associated with the human from the background;
perform annotation on the one or more identified body features to
generate annotation lines on each body feature corresponding to
body feature measurement locations utilizing a plurality of
annotation deep-learning networks that have been separately trained
on each body feature; generate body feature measurements from the
one or more annotated body features utilizing a sizing
machine-learning module based on the annotated body features and
the one or more user parameters; and generate body size
measurements by aggregating the body feature measurements for each
body feature.
Description
REFERENCE TO RELATED APPLICATIONS
[0001] This application is a non-provisional of and claims priority
from U.S. Ser. No. 62/809,218, filed on 22 Feb. 2019, and entitled
"SYSTEMS AND METHODS FOR FULL BODY MEASUREMENTS EXTRACTION USING
MULTIPLE DEEP LEARNING NETWORKS FOR BODY PARTS MEASUREMENTS."
[0002] This application is also a Continuation-In-Part (CIP) of PCT
Serial No. PCT/US19/27564, filed on 15 Apr. 2019, which itself is a
PCT of U.S. Ser. No. 16/195,802, filed on 19 Nov. 2018, issued as
U.S. Pat. No. 10,321,728, issued on 18 Jun. 2019, and entitled
"SYSTEMS AND METHODS FOR FULL BODY MEASUREMENTS EXTRACTION," which
itself claims priority from U.S. Ser. No. 62/660,377, filed on 20
Apr. 2018, and entitled "SYSTEMS AND METHODS FOR FULL BODY
MEASUREMENTS EXTRACTION USING A 2D PHONE CAMERA."
[0003] The entire disclosures of all referenced applications are
hereby incorporated by reference in their entireties herein.
NOTICE OF COPYRIGHTS AND TRADE DRESS
[0004] A portion of the disclosure of this patent document contains
material which is subject to copyright protection. This patent
document may show and/or describe matter which is or may become
trade dress of the owner. The copyright and trade dress owner has
no objection to the facsimile reproduction by anyone of the patent
disclosure as it appears in the U.S. Patent and Trademark Office
files or records, but otherwise reserves all copyright and trade
dress rights whatsoever.
FIELD OF THE INVENTION
[0005] Embodiments of the present invention are in the field of
automated body measurements and pertain particularly to extracting
body measurements of users using photos taken with a mobile
device.
BACKGROUND OF THE INVENTION
[0006] The statements in the background of the invention are
provided to assist with understanding the invention and its
applications and uses, and may not constitute prior art.
[0007] There are generally three approaches that have been tried to
generate or extract body measurements from images of users. A first
approach was to use 3D cameras that provide depth data, such as
MICROSOFT KINECT camera. With depth sensing, 3D body models can be
built to capture body sizing. However, not everyone has access to
3D cameras, and since there is no clear path to mass adoption at
the moment, it is not currently conceivable that such 3D cameras
become ubiquitous.
[0008] A second approach was to use 2D cameras to capture 2D
videos, and make use of 2D-to-3D reconstruction techniques to
recreate 3D body models to capture body sizing. Such a technique is
used by companies such as MTAILOR and 3DLOOK. In the 2D video
approach, a 3D body model is recreated, and the approach attempts
to perform a point cloud matching technique to match an existing 3D
body template with a pre-filled point cloud onto the newly created
3D body. However, the result may not be accurate when trying to fit
an existing template onto a unique user's 3D body. After the
matching of the template 3D body with the user's 3D body is
complete, sizing and measurements are obtained, but they are
generally not accurate.
[0009] A third approach was to use 2D cameras to capture 2D photos
instead of 2D videos, and similar to the previous method, utilize
2D-to-3D reconstruction techniques to capture body sizing. Such a
technique is used by AGISOFT, for example, a company that has
developed 3D reconstruction from 2D photos into 3D models. Using 2D
photos, instead of 2D videos, may involve photos captured at a
higher resolution, thus producing results with slightly higher
accuracy, but the other aforementioned problems remain.
[0010] In the existing methods using 2D videos or photos, 3D body
models are generated, and these approaches generally require the
user to have specific poses, stand at a specific distance from the
camera, in front of an empty background, wear tight fitting
clothing, and/or be partially nude wearing only underwear. Such
requirements for controlled environments and significant user
frictions are undesirable.
[0011] Therefore, it would be an advancement in the state of the
art to provide a system and method for accurately extracting body
measurements from 2D photos with 1) the user in any pose, 2) the
user standing in front of any background type, 3) the photos taken
at any distance, and 4) the user wearing any type of clothing, such
that everyone can easily take photos of themselves and benefit from
full body measurement extraction.
[0012] It is against this background that the present invention was
developed.
BRIEF SUMMARY OF THE INVENTION
[0013] The present invention relates to methods and systems for
extracting full body measurements using 2D user images, taken for
example from a mobile device camera.
[0014] More specifically, in various embodiments, the present
invention is a computer-implemented method for generating body size
measurements of a human, the computer-implemented method executable
by a hardware processor, the method comprising receiving one or
more user parameters from a user device; receiving at least one
image from the user device, the at least one image containing the
human and a background; performing body segmentation on the at
least one image to identify one or more body features associated
with the human from the background; performing annotation on the
one or more identified body features to generate annotation lines
on each body feature corresponding to body feature measurement
locations utilizing a plurality of annotation deep-learning
networks that have been separately trained on each body feature;
generating body feature measurements from the one or more annotated
body features utilizing a sizing machine-learning module based on
the annotated body features and the one or more user parameters;
and generating body size measurements by aggregating the body
feature measurements for each body feature.
[0015] In an embodiment, the performing body segmentation step
comprises generating a segmentation map of the body features on the
human; and cropping the one or more identified body features from
the human and the background based on the segmentation map, wherein
the cropped body features are used in the performing annotation
step.
[0016] In an embodiment, the performing body segmentation step
utilizes a segmentation deep-learning network that has been trained
on segmentation training data, wherein the segmentation training
data comprise one or more images for one or more sample humans and
a body feature segmentation for each body feature for the one or
more sample humans.
[0017] In an embodiment, the body feature segmentation is extracted
underneath clothing, and wherein the segmentation training data
comprises body segmentation estimating the sample human's body
underneath the clothing.
[0018] In an embodiment, the annotation deep-learning networks
utilize annotation training data comprising one or more sample
images for one or more sample humans and an annotation line for
each body feature for the one or more sample humans.
[0019] In an embodiment, the annotation line on each body feature
comprises one or more line segments corresponding to a given body
feature measurement, and the generating body feature measurements
step utilizes the annotation line on each body feature.
[0020] In an embodiment, the at least one image comprises at least
a front-view image and a side-view image of the human, and the
method further comprises the following steps after the performing
annotation step: calculating at least one circumference of at least
one annotated body feature utilizing line-annotated front-view and
side-view images and a height of the human; and generating the body
feature measurements from the at least one circumference utilizing
the sizing machine-learning module based on the at least one
circumference, the height, and the one or more user parameters.
[0021] In an embodiment, the sizing machine-learning module
comprises a random forest algorithm, and the sizing
machine-learning module is trained on ground truth data comprising
one or more sample body feature measurements for one or more sample
humans.
[0022] In an embodiment, the user parameters are selected from the
group consisting of a height, a weight, a gender, an age, and a
demographic information associated with the human.
[0023] In an embodiment, the receiving the one or more user
parameters from the user device comprises receiving user input of
the user parameters through the user device.
[0024] In an embodiment, the receiving the one or more user
parameters from the user device comprises receiving a measurement
performed by the user device.
[0025] In an embodiment, the at least one image is selected from
the group consisting of a front-view image of the human and a
side-view image of the human.
[0026] In an embodiment, the at least one image further comprises
an additional image of the human taken at an angle of approximately
45 degrees with respect to the front-view image of the human.
[0027] In an embodiment, the performing body segmentation step on
the at least one image further comprises receiving user input to
increase an accuracy of the body segmentation, and the user input
comprises a user selection of one or more portions of the
identified body features that correspond to a given region of the
human's body.
[0028] In an embodiment, the at least one image comprises at least
one image of a fully-clothed user or a partially-clothed user, and
the generating body feature measurements step further comprises
generating the body feature measurements on the at least one image
of the fully-clothed user or the partially-clothed user.
[0029] In an embodiment, the body size measurements comprise first
body size measurements, the method further comprises generating
second body size measurements using a second machine-learning
module, and an accuracy of the second body size measurements is
greater than an accuracy of the first body size measurements.
[0030] In an embodiment, the method further comprises determining
whether a given body feature measurement of the identified body
features corresponds to a confidence level below a predetermined
value; and in response to determining that the given body feature
measurement corresponds to a confidence level below the
predetermined value, performing 3D model matching using a 3D model
matching module on the identified body features to determine a
matching 3D model of the human, wherein one or more high-confidence
body feature measurements are used to guide the 3D model matching
module, performing body feature measurements based on the matching
3D model, and replacing the given body feature measurement with a
projected body feature measurement from the matching 3D model.
[0031] In an embodiment, the method further comprises determining
whether a given body feature measurement of the identified body
features corresponds to a confidence level below a predetermined
value; and in response to determining that the given body feature
measurement corresponds to a confidence level below the
predetermined value, performing skeleton detection using a skeleton
detection module on the identified body features to determine joint
positions of the human, wherein one or more high-confidence body
feature measurements are used to guide the skeleton detection
module, performing body feature measurement based on the determined
joint positions, and replacing the given body feature measurement
with a projected body feature measurement from the skeleton
detection module.
[0032] In an embodiment, the method further comprises
pre-processing the at least one image of the human and the
background before the performing body segmentation step, the
pre-processing comprises at least a perspective correction on the
at least one image, and the perspective correction is selected from
the group consisting of perspective correction utilizing a head of
the human, perspective correction utilizing a gyroscope of the user
device, and a perspective correction utilizing another sensor of
the user device.
[0033] In various embodiments, a computer program product is
disclosed. The computer program may be used for generating body
size measurements of a human, and may include a computer readable
storage medium having program instructions, or program code,
embodied therewith, the program instructions executable by a
processor to cause the processor to perform steps to receive one or
more user parameters from a user device; receive at least one image
from the user device, the at least one image containing the human
and a background; perform body segmentation on the at least one
image to identify one or more body features associated with the
human from the background; perform annotation on the one or more
identified body features to generate annotation lines on each body
feature corresponding to body feature measurement locations
utilizing a plurality of annotation deep-learning networks that
have been separately trained on each body feature; generate body
feature measurements from the one or more annotated body features
utilizing a sizing machine-learning module based on the annotated
body features and the one or more user parameters; and generate
body size measurements by aggregating the body feature measurements
for each body feature.
[0034] Yet another embodiment of the present invention is a
computer-implemented method for generating body sizing measurements
of a human, executable from a non-transitory computer readable
storage medium having program instructions embodied therein, the
program instructions executable by a processor to cause the
processor to receive one or more user parameters from a user
device; receive at least one image from the user device, the at
least one image containing the human and a background; perform body
segmentation on the at least one image to extract one or more body
features associated with the human from the background, the body
segmentation utilizing a segmentation deep-learning network that
has been trained on segmentation training data, wherein the
segmentation training data comprise one or more images for one or
more sample humans and a body feature segmentation for each body
feature for the one or more sample humans; perform body feature
annotation on the extracted body features for drawing an annotation
line on each body feature corresponding to a body feature
measurement, the body feature annotation utilizing a plurality of
annotation deep-learning networks that have been trained on
annotation training data unique to each body feature, wherein the
annotation training data comprise one or more images for one or
more sample body features and an annotation line for each body
feature for the one or more sample humans; generate body feature
measurements from the one or more annotated body features utilizing
a sizing machine-learning module based on the annotated body
features and the one or more user parameters; and generate body
size measurements by aggregating the body feature measurements for
each body feature. In one embodiment, the performing body
segmentation comprises generating a segmentation map of the body
features on the human; and cropping the one or more identified body
features from the human and the background.
[0035] In various embodiment, a system is described, including a
memory that stores computer-executable components; a hardware
processor, operably coupled to the memory, and that executes the
computer-executable components stored in the memory, wherein the
computer-executable components may include a components
communicatively coupled with the processor that execute the
aforementioned steps.
[0036] In another embodiment, the present invention is a
non-transitory, computer-readable storage medium storing executable
instructions, which when executed by a processor, causes the
processor to perform a process for generating body measurements,
the instructions causing the processor to perform the
aforementioned steps.
[0037] In another embodiment, the present invention is a system for
full body measurements extraction using a 2D phone camera, the
system comprising a user device having a 2D camera, a processor, a
display, a first memory; a server comprising a second memory and a
data repository; a telecommunications-link between said user device
and said server; and a plurality of computer codes embodied on said
first and second memory of said user-device and said server, said
plurality of computer codes which when executed causes said server
and said user-device to execute a process comprising the
aforementioned steps.
[0038] In yet another embodiment, the present invention is a
computerized server comprising at least one processor, memory, and
a plurality of computer codes embodied on said memory, said
plurality of computer codes which when executed causes said
processor to execute a process comprising the aforementioned
steps.
[0039] Other aspects and embodiments of the present invention
include the methods, processes, and algorithms comprising the steps
described herein, and also include the processes and modes of
operation of the systems and servers described herein.
[0040] Yet other aspects and embodiments of the present invention
will become apparent from the detailed description of the invention
when read in conjunction with the attached drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Embodiments of the present invention described herein are
exemplary, and not restrictive. Embodiments will now be described,
by way of examples, with reference to the accompanying drawings, in
which:
[0042] FIG. 1A shows an example flow diagram for body measurement
determination utilizing deep learning networks (DLNs) and machine
learning, in accordance with one embodiment of the invention.
[0043] FIG. 1B shows another example flow diagram for body
measurement determination using deep learning networks (DLNs) and
machine learning, in accordance with another embodiment of the
invention.
[0044] FIG. 1C shows a detailed flow diagram for body measurement
determination using deep learning networks (DLNs) and machine
learning, in accordance with another embodiment of the
invention.
[0045] FIG. 1D shows a detailed flow diagram for body feature
segmentation and annotation using deep learning networks (DLNs), in
accordance with one embodiment of the invention.
[0046] FIG. 1E shows an illustrative diagram for a machine learning
algorithm for body sizing determination from one or more feature
values obtained from the deep learning networks (DLNs), in
accordance with another embodiment of the invention.
[0047] FIGS. 1F and 1G show another illustrative flow diagram for
body size measurements utilizing deep learning networks (DLNs) and
machine learning, in accordance with another exemplary embodiment
of the invention.
[0048] FIG. 2 shows an example flow diagram for training the deep
learning networks (DLNs) and the machine learning module, which are
used with the flow diagram of FIG. 1A for body measurement
determination, in accordance with example embodiments of the
disclosure.
[0049] FIG. 3 shows an illustrative diagram of a user image (front
view) showing a human body wearing clothes captured for training
the segmentation and annotation DLNs.
[0050] FIG. 4 shows an illustrative diagram showing an annotator
segmenting one or more features of the human body under the
clothing from the background for training the segmentation DLN.
[0051] FIG. 5 shows an illustrative diagram of the body features of
the human body segmented from the background for training the
segmentation DLN.
[0052] FIG. 6 shows an illustrative diagram showing the annotator
annotating body annotation lines for training the annotation
DLN.
[0053] FIG. 7 shows another illustrative diagram of a user image
(side view) showing the annotator segmenting one or more body
features of the human body under the clothing for training the
segmentation DLN.
[0054] FIG. 8 shows another illustrative diagram (side view)
showing the annotator annotating body annotation lines for training
the annotation DLN.
[0055] FIG. 9 shows an illustrative client-server diagram for
implementing body measurement extraction, in accordance with one
embodiment of the invention.
[0056] FIG. 10 shows an example flow diagram for body measurement
determination (showing separate segmentation DLN, annotation DLN,
and sizing machine learning module), in accordance with one
embodiment of the invention.
[0057] FIG. 11 shows another example flow diagram for body
measurement determination (showing combined segmentation-annotation
DLN and sizing machine learning module), in accordance with another
embodiment of the invention.
[0058] FIG. 12 shows yet another example flow diagram for body
measurement determination (showing a combined sizing DLN), in
accordance with yet another embodiment of the invention.
[0059] FIG. 13 shows another example flow diagram for body
measurement determination (showing 3D human model and
skeleton-joint position model), in accordance with another
illustrative embodiment of the disclosure.
[0060] FIG. 14 shows an illustrative hardware architecture diagram
of a server for implementing one embodiment of the present
invention.
[0061] FIG. 15 shows an illustrative system architecture diagram
for implementing one embodiment of the present invention in a
client server environment.
[0062] FIG. 16 shows an illustrative diagram of a use case of the
present invention in which a single camera on a mobile device is
used to capture human body measurements, showing a front view of a
human in typical clothing standing against a normal background.
[0063] FIG. 17 shows an illustrative diagram of a mobile device
graphical user interface (GUI) showing user instructions for
capturing a front view photo, according to one embodiment of the
present invention.
[0064] FIG. 18 shows an illustrative diagram of the mobile device
GUI requesting the user to enter their height (and optionally other
user parameters, such as weight, age, sex, etc.) and to select
their preferred fit style (tight, regular, or loose fit), according
to one embodiment of the present invention.
[0065] FIG. 19 shows an illustrative diagram of the mobile device
GUI for capturing the front view photo, according to one embodiment
of the present invention.
[0066] FIG. 20 shows another illustrative diagram of the mobile
device GUI for capturing the front view photo, according to one
embodiment of the present invention.
[0067] FIG. 21 shows an illustrative diagram of the mobile device
GUI for capturing the side view photo, according to one embodiment
of the present invention.
[0068] FIG. 22 shows an illustrative diagram of the mobile device
GUI that is displayed while the system processes the captured
photos to extract the body measurements, according to one
embodiment of the present invention.
[0069] FIG. 23 shows an illustrative diagram of the mobile device
GUI showing a notification screen when the body measurements have
been successfully extracted, according to one embodiment of the
present invention.
DETAILED DESCRIPTION OF THE INVENTION
Overview
[0070] With reference to the figures provided, embodiments of the
present invention are now described in detail.
[0071] In the following description, for purposes of explanation,
numerous specific details are set forth in order to provide a
thorough understanding of the invention. It will be apparent,
however, to one skilled in the art that the invention can be
practiced without these specific details. In other instances,
structures, devices, activities, and methods are shown using
schematics, use cases, and/or flow diagrams in order to avoid
obscuring the invention. Although the following description
contains many specifics for the purposes of illustration, anyone
skilled in the art will appreciate that many variations and/or
alterations to suggested details are within the scope of the
present invention. Similarly, although many of the features of the
present invention are described in terms of each other, or in
conjunction with each other, one skilled in the art will appreciate
that many of these features can be provided independently of other
features. Accordingly, this description of the invention is set
forth without any loss of generality to, and without imposing
limitations upon, the invention.
[0072] Others have tried many different types of approaches to
generate or extract body measurements from images of users. All of
these approaches generally require the user to have specific poses,
stand at a specific distance from the camera, in front of an empty
background, wear tight fitting shirt, and/or go partially nude
wearing only underwear. Such requirements for controlled
environments and significant user friction are undesirable.
[0073] The present invention solves the aforementioned problems by
providing a system and method for accurately extracting body
measurements from 2D photos with 1) the human in any pose, 2) the
human standing in front of any background type, 3) the photos taken
at any distance, and 4) the human wearing any type of clothing,
such that everyone can easily take photos of themselves and benefit
from full body measurement extraction. Some embodiments of the
present invention do not involve any 3D reconstruction or 3D body
models, nor do they require specialized hardware cameras. Instead,
advanced computer vision combined with deep-learning techniques are
used to generate accurate body measurements no matter what the user
is wearing from photos provided from a simple mobile device camera.
In the present disclosure, the term "2D phone camera" is used to
represent any cameras embedded in, or connected to, computing
devices, such as smart phones, tablets, laptops, or desktops.
Multiple Deep Learning Networks for Body Feature Measurements
[0074] FIG. 1A shows a diagram of an example flow for body
measurement determination operations, in accordance with example
embodiments of the disclosure. In some embodiments of the prevent
invention, computer vision techniques and deep learning are applied
to one front view photo and one side view photo of the user, plus
the user's height, and possibly other user parameters such as
weight, sex, age, etc. and generate full body measurements using
one or more deep learning networks that have been trained on
annotated body measurements collected and annotated for sample
humans. As more data is collected by the system, the accuracy of
the body measurements automatically improves. In some other
embodiments, perspective correction, human background subtraction,
skeleton detection, and 3D model matching approaches, utilizing
computer vision techniques, are used to improve on any
low-confidence body measurements from the deep learning approach.
This hybrid approach significantly improves body measurement
accuracy, and increases user satisfaction with the body
measurements. In case the body measurements are used for custom
garment manufacture, the resultant accuracy improves customer
service and reduces return rates for the manufactured custom
garments.
[0075] The overall process 100 starts at step 102, where
normalization data (one or more user parameters), such as a height
of the user, is obtained, generated, and/or measured in order to
perform a normalization or a scaling. In another embodiment, a
weight may also be used in conjunction with the height. Both user
parameters may be determined automatically (e.g., using computer
vision algorithms or mined from one or more databases), or
determined from the user (e.g., user input). In one embodiment,
from these user parameters, a body mass index (BMI) may be
calculated. The BMI may be used to calibrate the body measurement
extraction using both the body weight and height. Additional user
parameters may include at least one of a height, a weight, a
gender, an age, race, country of origin, athleticism, and/or other
demographic information associated with the user, among others. The
height of the user is used to normalize, or scale, front and/or
side-view photos and provide a reference for a human in the photo.
The other user parameters, such as the weight, BMI index, age, sex,
and so forth, are used as additional inputs into the system to
optimize the body sizing measurements. In one embodiment, the other
user parameters may also be obtained automatically from the user
device, from one or more third-party data sources, or from the
server.
[0076] At step 104, one or more user photos may be received; for
example, at least one front and/or side view photos of a given user
may be received. In another embodiment, the photos may be obtained
from the user device (e.g., mobile phone, laptop, tablet, etc.). In
another embodiment, the photos may be obtained from a database
(e.g., a social media database). In another embodiment, the user
photos include a photo showing a front view and a photo showing a
side view of the entire body of the user. In some embodiments, only
one photo, such as a front view, is utilized and the one photo is
sufficient to perform accurate body measurement extraction. In yet
other embodiments, three or more photos are utilized, including in
some embodiments a front view photo, a side view photo, and a photo
taken at an approximately 45 degree angle. Other combinations of
user photos are within the scope of the present invention, as would
be recognized by one of ordinary skill in the art. In some
embodiments, a user video, for example a front view, a 90, 180, or
even 360 degree view of the user may be received. From the user
video, one or more still frames or photos, such as a front view, a
side view, and/or a 45-degree view of the user are extracted from
the video, and used in the process that follows. Steps 102 and 104
may be performed in any order in various embodiments of the present
invention, or the two steps may be implemented in parallel.
[0077] In one embodiment, the system may automatically calculate
(e.g., using one or more AI-algorithms) body measurements using the
photos and the normalization data, as further described below in
connection with the following steps. In another embodiment, the
user may indicate whether the user is dressed in tight, normal, or
loose clothing for more accurate results.
[0078] In one embodiment, the images may be taken at a specified
distance (e.g., approximately 10 feet away from the camera of a
user's device). In one embodiment, the images may be taken with the
user having a specific pose (e.g., arms in a predetermined
position, legs spread at a shoulder length, back straight,
"A-pose," etc.). In another embodiment, multiple images of a given
position (e.g., front and side view photos) may be taken and an
average image may be determined for each position. This may be
performed to increase accuracy. In another embodiment, the user may
be positioned against a background of a specific type (e.g., a
neutral color, or having a predetermined background image). In some
embodiments, the user may be positions against any type of
background. In one embodiment, the front and side view photos may
be taken under similar lighting conditions (e.g., a given
brightness, shadow, and the like). In another embodiment, the front
and side view photos may include images of the user wearing
normally fitted clothing (e.g., not extra loose or extra tight).
Alternatively, or additionally, the front and side view photos may
include images of the user partially clothed (e.g., shirtless), or
having a different type of fit (e.g., tight, loose, etc.) depending
on the needs of the AI-based algorithms and associated
processes.
[0079] In some embodiments, a pre-processing on the one or more
photos of the user (not shown in FIG. 1A), such as a perspective
correction, may be performed on the front and side view photos, if
needed. For example, the system may use OpenCV, an open-source
machine vision library, and may make use of features of the head in
the front and side view photos and the user's height as references
for perspective correction. In this way, embodiments of the
disclosure may avoid determining measurements which are inaccurate
as far as the proportions of the lengths of the body go, such as
torso length and leg length. Optionally, a perspective side photo
showing where the camera is positioned relative to the person being
photographed may yield even more accurate perspective correction by
allowing the system to calculate the distance between the camera
and the user. In some embodiments, the system may instead use
gyroscope data provided by the user device (or a peripheral device
connected to the user device, such as an attached computer device)
to detect a photo perspective angle, and perform perspective
correction based on this photo perspective angle.
[0080] In some embodiments, one or more additional pre-processing
steps (not shown in FIG. 1A) may be performed on the one or more
photos of the user. Various computer vision techniques may be
utilized to further pre-process the one or more images. Examples of
pre-processing steps may include, in addition to perspective
correction, contrast, lighting, and other image processing
techniques to improve the quality of the one or more images before
further processing.
[0081] At step 105, a body feature, such as a body part of the
human (e.g., a neck, an arm, a leg, etc.), may be segmented from
the image using a first deep learning network (DLN) known as a
segmentation DLN. In one embodiment, at step 105, a segmentation
map is generated which segments the human in the image into one or
more body features. In one embodiment, "deep learning" may refer to
a class of machine learning algorithms that use a cascade of
multiple layers of nonlinear processing units for feature
extraction and transformation modeled after neural networks. In one
embodiment, the successive layers may use the output from the
previous layer as input. In one embodiment, the "deep" in "deep
learning" may refer to the number of layers through which the data
is transformed. An example of body feature segmentation is
explained and shown in reference to FIGS. 3-5 below.
[0082] Before performing this segmentation step on data from a real
user, the system may have been trained first, for example, on
sample photos of humans posing in different environments in
different clothing, for example, with hands at 45 degrees,
sometimes known as the "A-pose", as described in relation to FIG.
2. In some embodiments, any suitable deep learning architecture may
be used, such as deep neural networks, deep belief networks, and/or
recurrent neural networks. In another embodiment, the deep learning
algorithms may learn in supervised (e.g., classification) and/or
unsupervised (e.g., pattern analysis) manners. Further, the deep
learning algorithms may learn multiple levels of representations
that correspond to different levels of abstraction of the
information encoded in the images (e.g., body, body part, etc.). In
another embodiment, the images (e.g., the front and side photos)
may be represented as a matrix of pixels. For example, in one
embodiment, the first representational layer may abstract the
pixels and encode edges; the second layer may compose and encode
arrangements of edges; the third layer may encode a nose and eyes;
and the fourth layer may recognize that the image contains a face,
and so on.
[0083] In one embodiment, the segmentation DLN algorithm may be
trained with segmentation training data, as described in relation
to FIG. 2 below. In some embodiments, the segmentation training
data may include sample humans with segmented body features. The
segmentation may be performed on the sample humans either manually
or automatically. In some embodiments, the training data includes
medical data, for example from CAT scans, MRI scans, and so forth.
In some embodiments, the training data includes data from previous
tailor or 3D body measurements that include 3D body scans from 3D
body scanners and "ground truth" data. In some embodiments, the 3D
body scans may be used to extract approximate front and/or side
view photos, in cases where the front and side view photos are not
explicitly available. In some embodiments, the ground truth data
comprises human tailor-measured data; while in other embodiments,
the ground truth data comprises automatically extracted 1D body
size measurements from the 3D body scans. In some embodiments, 3D
body scan data from the "SizeUSA" data set, which is a commercial
sample of 3D body scans obtained on about 10,000 human subjects
(both male and female), may be utilized. In other embodiments, 3D
body scan data from the "CAESAR" data set may be utilized, which is
another commercial sample of 3D body scans obtained on about 4,000
human subjects, and also includes manually-measured ground truth
data using a human tailor. In yet other embodiments, an
organization utilizing the present invention may capture their own
front and side photos, along with suitable ground truth data using
a human tailor, for training the segmentation DLN.
[0084] In yet other embodiments, the segmentation training data may
have been automatically generated, for example as described in
related U.S. Ser. No. 16/517,391, filed on 19 Jul. 2019, and
entitled "METHODS AND SYSTEMS FOR AUTOMATIC GENERATION OF MASSIVE
TRAINING DATA SETS FROM 3D MODELS FOR TRAINING DEEP LEARNING
NETWORKS," which is incorporated by reference herein in its
entirety as if fully set forth herein.
[0085] At step 106, the identified body features may be segmented,
separated, and/or cropped from the rest of the human and the
background using the segmentation map generated in step 105. The
segmentation/cropping may be actual or virtual cropping. The part
of the image corresponding to each identified body feature may be
cropped, segmented, or separated from the rest of the image, and
that part of the image passed to the annotation step 107. By
segmenting or cropping the identified body features from the rest
of the image, the DLN used in annotation step 107 can be specially
or separately trained on each separate body feature, increasing
both accuracy and reliability.
[0086] At step 107, an annotation line for each body feature that
was segmented or cropped at step 106 may be drawn using a plurality
of deep learning networks (DLNs), for example a plurality of
annotation DLNs. In one embodiment, there is one body feature
annotation DLN for the entire body. In one preferred embodiment,
there is a separate body feature annotation DLN for each body
feature. An advantage of using a separate body feature annotation
DLN for each body feature is increased accuracy and reliability in
body feature measurements. Each body feature DLN may be separately
trained on separate and unique data for each body feature. The
specificity of data on each body feature increases the accuracy and
reliability of the DLN, and also increases the speed of convergence
of the neural network layer training. An example of body feature
annotation is explained and shown in reference to FIGS. 6-8
below.
[0087] In one embodiment, the system may generate and extract body
feature measurements by using an AI-based algorithm such as an
annotation DLN, for example, by first drawing annotation lines from
signals obtained from the body features. Each annotation line may
be different for each body feature and may be drawn differently.
For example, for the bicep width or circumference, the system may
draw a line perpendicular to the skeletal line at the bicep
location; for the chest, the system may connect two chest dots
instead. From the annotation of each body feature, a body feature
measurement may then be obtained by normalizing on the user's
height received in step 102, as described further below.
[0088] Before performing this annotation step on data from a real
user, the system may have been trained first, for example, on
photos of sample humans posing in different environments in
different clothing with annotated body features, as described in
relation to FIG. 2 further below. The annotation lines may be drawn
manually or automatically on the sample humans. The segmentation
and annotation DLNs are described in more detail in relation to
FIGS. 1B, 1C, and 1D.
[0089] In yet other embodiments, the annotation training data may
have been automatically generated, for example as described in
related U.S. Ser. No. 16/517,391, filed on 19 Jul. 2019, cited
above.
[0090] At step 108, a body feature measurement may be estimated for
each body feature that had an annotation line drawn at step 107
using one or more machine learning algorithms, for example a sizing
machine learning (ML) algorithm. In one embodiment, the sizing ML
algorithm comprises a random forest machine learning module. In one
embodiment, there is a separate sizing ML module for each body
feature. In some embodiments, there is one sizing ML module for the
entire body. In one embodiment, the system may determine the sizes
of the body features using as input the height received in step 102
to normalize the sizing estimates. In order to do this, the
annotation DLN in one embodiment draws a "full body" annotation
line indicating a location of the subject's height, with a dot
representing a bottom of the subject's feet and another dot
representing a top of the subject's head. This "fully body"
annotation line is used to normalize other annotation lines by the
subject's known height provided in step 102. In other words, the
height of the subject in the image is detected, and used along with
the known actual height to normalize all annotation line
measurements. This process may be thought of as "height reference
normalization," using the subject's known height as a standard
measurement for normalization.
[0091] In another embodiment, additional user demographic data,
such as, but not limited to, weight, a BMI index, a gender, an age,
and/or other demographic information associated with the user
received in step 102 is used as input to the sizing ML algorithm
(such as random forest), described in greater detail in relation to
FIG. 1E.
[0092] The system may also use other algorithms, means, and medians
for each body feature measurement. The annotation DLN and sizing ML
may be implemented as one sizing DLN, that annotates and performs
measurements on each body feature, or may be implemented as two
separate modules, an annotation DLN that annotates each body
feature, and a separate sizing ML module that performs the
measurements on the annotated body feature. Similarly, various
alternative architectures for implementing the segmentation DLN of
step 106, the annotation DLN of step 107, and the sizing ML module
of step 108 are described in relation to FIGS. 10-12 below. For
example, FIG. 10 corresponds to the architecture shown in FIG. 1A,
in which the segmentation DLN, annotation DLN, and sizing ML module
are separate modules. In contrast, FIG. 11 corresponds to an
alternative architecture (not shown in FIG. 1A) in which the
segmentation DLN and annotation DLN are combined into a single
annotation DLN (that effectively performs both segmentation and
annotation) followed by a sizing ML module. Finally, FIG. 12
corresponds to yet another alternative architecture (not shown in
FIG. 1A) in which the segmentation DLN, annotation DLN, and sizing
ML module are all combined into a single sizing DLN that
effectively performs all functions of segmentation, annotation, and
size measurement.
[0093] At step 110, a confidence level for each body feature
measurement may be determined, obtained, or received from the
sizing ML module from step 108. In addition to outputting the
predicted body measurements for each body feature, the sizing ML
module also outputs a confidence level for each predicted body
feature measurement, which is then utilized to determine if any
other approaches should be utilized to improve on the output, as
described below. In another embodiment, the confidence level may be
based on a confidence interval. In particular, a confidence
interval may refer to a type of interval estimate, computed from
the statistics of the observed data (e.g., the front and side
photos encoding image data), that might contain the true value of
an unknown population parameter (e.g., a measurement of a body
part). The interval may have an associated confidence level that
may quantify the level of confidence that the parameter lies in the
interval. More strictly speaking, the confidence level represents
the frequency (i.e. the proportion) of possible confidence
intervals that contain the true value of the unknown population
parameter. In other words, if confidence intervals are constructed
using a given confidence level from an infinite number of
independent sample statistics, the proportion of those intervals
that contain the true value of the parameter will be equal to the
confidence level. In another embodiment, the confidence level may
be designated prior to examining the data (e.g., the images and
extracted measurements therefrom). In one embodiment, a 95%
confidence level is used. However, other confidence levels can be
used, for example, 90%, 99%, 99.5%, and so on.
[0094] In various embodiments, a confidence interval and
corresponding confidence level may be determined based on a
determination of a validity and/or an optimality. In another
embodiment, validity may refer to the confidence level of the
confidence interval holding, either exactly or to a good
approximation. In one embodiment, the optimality may refer to a
rule for constructing the confidence interval should make as much
use of the information in the dataset (images and extracted
features and measurements) as possible.
[0095] At step 112, it may be determined whether the confidence
level is greater than a predetermined value. If it is determined
that the confidence level is greater than the predetermined value,
then the process may proceed to step 114, where the high-confidence
body feature measurements may be outputted. If it is determined
that the confidence level is less than the predetermined value,
then the process may proceed to step 116 or step 118. The steps 116
and 118 are illustrative of one or more, optional, fallback
algorithms for predicting or projecting estimated body feature
measurements for those body features for which the deep-learning
approach has a low confidence. Together with the high-confidence
body feature measurements from the deep-learning approach (shown in
dashed lines), and the projected body feature measurements from the
alternative fallback algorithms for the low-confidence body feature
measurements, are later synthesized into a complete set of
high-confidence body feature measurements as described below. As
noted, in another embodiment, the confidence level may be
designated prior to examining the data (e.g., the images and
extracted measurements therefrom).
[0096] In particular, at steps 116 and 118, other optional models
(e.g., AI-based or computer vision-based models) may be applied. At
step 116, and according to one optional embodiment, a 3D human
model matching algorithm may be applied. For example, the system
may first utilize OpenCV and/or deep learning techniques to extract
the human body from the background. The extracted human body is
then matched to one or more known 3D human models in order to
obtain body feature measurements. Using this technique and a
database of existing 3D body scans, for example a database of
several thousand 3D body scans, the system may match the closest
body detected with the 3D body scans' points. Using the closest
matching 3D model, the system may then extract body feature
measurements from the 3D model. This technique is described in more
detail in relation to FIG. 13 below.
[0097] Alternatively, and/or additionally, at step 118, other
models, such as a skeleton/joint position model may be applied. In
one embodiment, skeleton/joint detection may be performed using
OpenPose (discussed further below), an open source algorithm for
pose detection. Using this technique to obtain the skeleton and
joint positions, the system may then draw lines between the
appropriate points, using an additional deep learning network (DLN)
if necessary, that indicates positions of the middle of the bone
that are drawn on top of the user photos to indicate various key
skeletal structures, showing where various body parts, such as the
shoulders, neck, and arms are. From this information, body feature
measurements may be obtained from the appropriate lines. For
example, a line connecting the shoulder and the wrist may be used
to determine the arm length. This technique is described in detail
in relation to FIG. 13 below.
[0098] In one embodiment, the 3D model algorithm and the
skeleton/joint position models are combined as follows (though this
is not shown explicitly in FIG. 1A). Using a database of existing
3D body scans, for example a database of several thousand 3D body
scans, the system may match the closest skeleton detection with the
3D body scans' skeleton points, showing points and lines that
indicate positions of the bone that indicate various key skeletal
structures, showing where various body parts, such as the
shoulders, neck, and arms are. Once the closest matching 3D model
is matched, the system may extract body feature measurements from
the 3D model.
[0099] In either or both cases, at step 120, or at step 122, or
both, the high-confidence body feature measurements may be
projected (e.g., estimated). In particular, the estimate of the
high-confidence body feature measurement may be performed using a
different process from the first, lower-confidence deep-learning
process (e.g., that shown and described in connection with step
108, above).
[0100] One advantageous feature of this approach is that the
high-confidence body feature measurements from step 114 (shown as a
dashed line) may be used as inputs to assist with calibrating the
other models, for example the 3D human model algorithm in step 116
and the skeleton/joint position model in step 118. That is, the
high-confidence body feature measurements from the deep learning
approach obtained in step 108, may be used to assist the other
models, for example 3D human model 116 and/or skeleton/joint
position model 118. The other models (116 and/or 118) may then be
used to obtain projected high-confidence body feature measurements
for those body feature measurements that were determined to be have
a confidence below a predetermined value in step 112. Later, the
projected high-confidence body feature measurements may replace or
supplement the low-confidence body feature measurements from the
deep-learning approach.
[0101] Further, at step 124, the high confidence body feature
measurements determined at step 120 and/or step 122 may be used to
determine a high-confidence body feature measurement. In such a
way, various models, that is, the 3D human model and the
skeleton/joint position model, may both be used to further improve
the accuracy of the body feature measurements obtained in step 114.
Therefore, the high-confidence body feature measurements are
aggregated--the high-confidence body feature measurements from step
114 (e.g., the deep-learning approach) are combined with the
projected high-confidence body feature measurements from steps 120
and 122 (e.g., the other models).
[0102] At step 126, the high-confidence body feature measurements
are aggregated into complete body measurements of the entire human
body, and are then output for use. In particular, the body
measurements may be output to a user device and/or a corresponding
server, for example associated with a company that manufactures
clothing based on the measurements. In one embodiment, the output
may be in the form of a text message, an email, a textual
description on a mobile application or website, combinations
thereof, and the like. The complete body measurements may then be
used for any purposes, including but not limited to custom garment
generation. One of ordinary skill in the art would recognize that
the output of the complete body measurements may be utilized for
any purpose in which accurate and simple body measurements are
useful, such as but not limited to fitness, health, shopping, and
so forth.
[0103] FIG. 1B shows another example flow diagram 150 for body
measurement determination using deep learning networks (DLNs) and
machine learning, in accordance with another embodiment of the
invention. In step 151, input data 152, which comprises a front
photo, side photo, and user parameters (height, weight, age, sex,
etc.) are received. In step 153, one or more image processing steps
are applied. First, optional image pre-processing (perspective
correction, human cropping, resizing etc.) steps may be performed.
Next, the body features are cropped from the human and the
background using a segmentation map generated by a segmentation
deep-learning network (DLN) 154. Then, multiple annotation DLNs 155
are applied to the cropped body features for annotating each body
feature, as described in greater detail in relation to FIG. 1D.
Finally, the sizing machine learning module (ML) 156 is applied to
the annotated body features for determining the body size
measurements from the annotation lines and one or more of the user
parameters, as described in greater detail in relation to FIG. 1E.
Finally, in step 155, the body size measurements (for example, 16
standard body part sizes) are output, shown illustratively as
output data 158. The output 158 may include the sizing result (a
set of standard body size measurements, such as neck, shoulder,
sleeve, height, outseam, inseam, etc.), and may also include the
front and side photos annotated with the annotation lines.
[0104] FIG. 1C shows a detailed flow diagram 160 for body
measurement determination using deep learning networks (DLNs) and
machine learning, in accordance with another embodiment of the
invention. Inputs to the body measurement process include front
photo 161, side photo 162, height 163, and other user parameters
(weight, age, sex, etc.) 164. The front photo 161 is pre-processed
in step 165, while the side photo 162 is pre-processed in step 166.
Examples of pre-processing steps, such as perspective correction,
human cropping, image resizing, etc. were previously discussed. At
step 167, the pre-processed front photo is used as input to DLN 1
(described in more detail in relation to FIG. 1D) to extract
annotation lines for the front photo 161. At step 168, the
pre-processed side photo is used as input to DLN 2 to analogously
extract annotation lines for the side photo 161. The annotation
lines for each body part from the front view 169 are output from
DLN 1 and the annotation lines for each body part from the side
view 170 are output from DLN 2. At step 171, the two sets of
annotation lines from the front photo 161 and the side photo 162
are utilized along with the height normalization reference 175
received from height input 163 to calculate a circumference of each
body part. At step 172, the circumference of each body part, along
with the height and other user parameters 176 received from inputs
163 and 164 are utilized in a machine learning algorithm, such as
random forest (described in more detail in relation to FIG. 1E), to
calculate one or more body size measurements. At step 173, the body
size measurement results (length of each standard measurement) are
output. Finally, the body measurement process ends at step 174.
Illustrative Deep Learning Network and Machine Learning
Architectures
[0105] FIG. 1D shows a detailed flow diagram 180 for body feature
segmentation and annotation, in accordance with one embodiment of
the invention. In one embodiment, the body feature segmentation and
annotation is done using a deep learning network (DLN) using
training data as described above. In one embodiment, the body part
segmentation and annotation is performed using a convolutional
neural network (CNN) combined with a pyramid scene parsing network
(PSPNet) for improved global and local context information. In a
PSPNet, the process may utilize "global & local context
information" from different sized regions that are aggregated
through a "pyramid pooling module." As shown in FIG. 1D, the input
image 181 is first passed through a convolutional neural network
(CNN) 182 to obtain a feature map 183 which classifies or segments
each pixel into a given body part and/or annotation line. Next,
global & local context information is extracted from the
feature map utilizing the pyramid pooling module 184, which
aggregates information from the image on different size scales.
Finally, the data is passed through a final convolution layer 185
to classify each pixel into body feature segments and/or annotation
lines 186.
[0106] In greater detail, from an input image 181, a CNN 182 is
first used to obtain a feature map 183, then a pyramid pooling
module 184 is used to extract different sub-regions' features;
followed by up-sampling and concatenation layers to form the final
feature representation, which carries both local and global context
information. Finally, the feature representation is fed to a final
convolution layer 185 to obtain the final per-pixel prediction. In
the example shown in FIG. 1D, the pyramid pooling module 184
combines features under four different scales. The largest scale is
global. The subsequent levels separate the feature map into
different sub-regions. The output of different levels in the
pyramid pooling module 184 comprise the feature map under different
scales. In one embodiment, to maintain the weight of the global
features, a convolution layer may be used after each pyramid level
to reduce the dimension of context representation, as shown in FIG.
1D. Next, the low-dimension feature maps are up-sampled to get the
same size features as the original feature map. Finally, the
different feature levels are concatenated with the original feature
map 183 for the pyramid pooling module 184 output. In one
embodiment, by using a four-level pyramid, as shown, the pooling
windows cover the whole, half, and smaller portions of the original
image 181.
[0107] In one embodiment, the PSPNet algorithm is implementation as
described in Hengshuang Zhao, et al., "Pyramid Scene Parsing
Network," CVPR 2017, Dec. 4, 2016, available at arXiv:1612.01105,
which is hereby incorporated by reference in its entirety herein as
if fully set forth herein. PSPNet is only one illustrative deep
learning network algorithm that is within the scope of the present
invention, and the present invention is not limited to the use of
PSPNet. Other deep learning algorithms are also within the scope of
the present invention. For example, in one embodiment of the
present invention, a convolutional neural network (CNN) is utilized
to extract the body segments (segmentation), and a separate CNN is
used to annotate each body segment (annotation).
[0108] FIG. 1E shows an illustrative diagram 190 for a machine
learning algorithm for body measurement determination from one or
more feature values 191 obtained from the deep learning networks
(DLNs), in accordance with another embodiment of the invention. In
one embodiment, the body part sizing is determined using a random
forest algorithm, a specialized machine learning algorithm. Random
forests use a multitude of decision tree predictors, such that each
decision tree depends on the values of a random subset of the
training data, which minimizes the chances of "overfitting". In one
embodiment, the random forest algorithm is implementation as
described in Leo Breiman, "Random Forests," Machine Learning, 45,
5-32, 2001, Kluwer Academic Publishers, Netherlands, Available at
doi.org/10.1023/A:1010933404324, which is hereby incorporated by
reference in its entirety herein as if fully set forth herein.
Random forest is only one illustrative machine learning algorithm
that is within the scope of the present invention, and the present
invention is not limited to the use of random forest. Other machine
learning algorithms, including but not limited to, nearest
neighbor, decision trees, support vector machines (SVM), Adaboost,
Bayesian networks, various neural networks including deep learning
networks, evolutionary algorithms, and so forth, are within the
scope of the present invention. The input to the machine learning
algorithm are the features values (x) 191, which comprise the
circumferences of the body parts obtained from the deep-learning
networks, the height, and the other user parameters, as described
in relation to FIG. 1C. The output of the machine learning
algorithm are the predicted values for the sizing measurements (y)
192.
[0109] As noted, embodiments of devices and systems (and their
various components) described herein can employ artificial
intelligence (AI) to facilitate automating one or more features
described herein (e.g., providing body extraction, body
segmentation, measurement extraction, and the like). The components
can employ various AI-based schemes for carrying out various
embodiments/examples disclosed herein. To provide for or aid in the
numerous determinations (e.g., determine, ascertain, infer,
calculate, predict, prognose, estimate, derive, forecast, detect,
compute) described herein, components described herein can examine
the entirety or a subset of the data to which it is granted access
and can provide for reasoning about or determine states of the
system, environment, etc. from a set of observations as captured
via events and/or data. Determinations can be employed to identify
a specific context or action, or can generate a probability
distribution over states, for example. The determinations can be
probabilistic--that is, the computation of a probability
distribution over states of interest based on a consideration of
data and events. Determinations can also refer to techniques
employed for composing higher-level events from a set of events
and/or data.
[0110] Such determinations can result in the construction of new
events or actions from a set of observed events and/or stored event
data, whether the events are correlated in close temporal
proximity, and whether the events and data come from one or several
event and data sources. Components disclosed herein can employ
various classification (explicitly trained (e.g., via training
data) as well as implicitly trained (e.g., via observing behavior,
preferences, historical information, receiving extrinsic
information, etc.)) schemes and/or systems (e.g., support vector
machines, neural networks, expert systems, Bayesian belief
networks, fuzzy logic, data fusion engines, etc.) in connection
with performing automatic and/or determined action in connection
with the claimed subject matter. Thus, classification schemes
and/or systems can be used to automatically learn and perform a
number of functions, actions, and/or determinations.
[0111] A classifier may map an input attribute vector, z=(z1, z2,
z3, z4, . . . , zn), to a confidence that the input belongs to a
class, as by f(z)=confidence(class). Such classification may employ
a probabilistic and/or statistical-based analysis (e.g., factoring
into the analysis utilities and costs) to determinate an action to
be automatically performed. Another example of a classifier that
can be employed is a support vector machine (SVM). The SVM operates
by finding a hyper-surface in the space of possible inputs, where
the hyper-surface attempts to split the triggering criteria from
the non-triggering events. Intuitively, this makes the
classification correct for testing data that is near, but not
identical to training data. Other directed and undirected model
classification approaches include, e.g., naive Bayes, Bayesian
networks, decision trees, neural networks, fuzzy logic models,
and/or probabilistic classification models providing different
patterns of independence can be employed. Classification as used
herein also is inclusive of statistical regression that is utilized
to develop models of priority.
[0112] Illustrative Body Feature Measurement Estimation Process
FIGS. 1F and 1G show an illustrative flow diagram 193 for body size
measurements utilizing deep learning networks (DLNs) and machine
learning, in accordance with another exemplary embodiment of the
invention. An estimation request is received at step 194 from the
user device, including front and side photos, and one or more user
parameters (e.g., height, weight, age, sex, etc.) from an IOS SDK,
a JAVASCRIPT SDK, or an ANDROID SDK, for example. Deep learning
automatic segmentation and annotation is then performed at step
195, as described in greater detail above. In one illustrative
embodiment and as shown in FIG. 1F, deep learning pre-processing is
performed to crop around the target body features, resize, and
normalize the image to a color range in [-1, +1]. Next, deep
learning annotation is performed on the body features to generate
Gaussian heat maps around keypoints, as described in detail above.
Finally, deep learning post processing is performed to convert the
Gaussian heat maps around keypoints to the keypoints' 2D
coordinates by finding maximum values and converting the 2D
coordinates to original coordinates.
[0113] The process in FIG. 1F is continued in FIG. 1G. At step 196,
a confidence check is performed to check the keypoints'
relationships based on crossing angles. If the confidence check is
not successful, at step 197, a manual annotation is performed by an
annotator to fix keypoints using an annotation GUI, as described
above. If the manual annotation is successful, this step 197
results in manual front and side annotation keypoints.
[0114] If the confidence check at step 196 or the manual annotation
at step 197 is successful, the resultant keypoints are used in the
machine learning sizing estimation module at step 198. As described
above, at step 198, machine learning pre-processing converts pixel
coordinates to cm real-world coordinates based on user height, and
the computes real-world lengths between related keypoints. Finally,
machine learning size estimation is performed using a random forest
algorithm using as input the length information and the one or more
user parameters, as described in detail above.
[0115] Finally, at step 199, if the machine learning size
estimation at step 198 has succeeded, the size estimation result is
provided as output. If the machine learning size estimation at step
198 has failed, or the automatic and/or manual annotation has
failed at steps 196 or 197, and error code and error reason is
provided as output at step 199.
Training the Deep Learning Networks and Machine Learning
Modules
[0116] FIG. 2 shows a diagram 200 of an exemplary flow diagram for
training the segmentation DLN, the annotation DLN, and the sizing
ML, which are utilized in generating body measurements, in
accordance with example embodiments of the present invention. At
step 202, one or more photos are received. For example, front and
side view photos of a given user may be received. In another
embodiment, the photos may be obtained from the user device (e.g.,
mobile phone, laptop, tablet, etc.). In another embodiment, the
photos may be obtained from a database (e.g., a social media
database). In another embodiment, the photos from the user include
a photo showing a front view and a photo showing a side view of the
entire body of the user.
[0117] As noted, in one embodiment, the images may be taken at a
specified distance (e.g., approximately 10 feet away from the
camera of a user's device). In one embodiment, the images may be
taken with the user having a specific pose (e.g., arms in a
predetermined position, legs spread at a shoulder length, back
straight, "A-pose," etc.). In another embodiment, multiple images
of a given position (e.g., front and side view photos) may be taken
and an average image may be determined for each position. This may
be performed to increase accuracy. In another embodiment, the user
may be positioned against a background of a specific type (e.g., a
neutral color, or having a predetermined background image). In
another embodiment, the front and side view photos may be taken
under similar lighting conditions (e.g., a given brightness,
shadow, and the like). In another embodiment, the front and side
view photos may include images of the user wearing normally fitted
clothing (e.g., not extra loose or extra tight). Alternatively,
and/or additionally, the front and side view photos may include
images of the user partially clothed (e.g., shirtless), or having a
different type of fit (e.g., tight, loose, etc.) depending on the
needs of the AI-based algorithms and associated processes.
[0118] In some embodiments, one or more pre-processing steps (not
shown in FIG. 2), such as a perspective correction may be performed
on the front and side view photos, if needed. For example, the
system may use OpenCV, an open source machine vision library, and
may make use of features of the head in the front and side view
photographs and the user's height as references for perspective
correction. In this way, embodiments of the disclosure may avoid
determining measurements which are inaccurate as far as the
proportions of the lengths of the body go, such as torso length and
leg length. Optionally, a perspective side photo showing where the
camera is positioned relative to the person being photographed may
yield even more accurate perspective correction by allowing the
system to calculate the distance between the camera and the user.
In some embodiments, the system may instead use gyroscope data
provided by the user device (or a peripheral device connected to
the user device, such as an attached computer device) to detect a
photo perspective angle, and perform perspective correction based
on this photo perspective angle. Other pre-processing steps, such
as contrast, lighting, or other image processing techniques may be
utilized to pre-process the received images in order to facilitate
the following steps.
[0119] At step 204, an annotator may segment body features, such as
body parts, under the clothing using human intuition. In one
embodiment, the body parts may be color-coded for convenience. In
particular, body segmentation may be performed by a human to
extract a human body, excluding clothing, from a background of the
photos. For example, the human annotator may be used to visually
edit (e.g., trace out and color code) photos and indicate which
body parts correspond to which portions of the photos to extract
the human body, excluding clothing, from the background. In one
embodiment, the photos may include humans posing in different
environments in different clothing, with hands at 45 degrees
("A-pose"). As noted, accurate body outlines may be drawn by human
annotators manually from the background. The ability of human
annotators to determine the body shape of photographed humans under
any kind of clothing, especially by skilled annotators who are
experienced and who can provide accurate and reliable body shape
annotations, ensures high performance for the system. The body
outlines may be drawn on any suitable software platform, and may
use a peripheral device (e.g., a smart pen) for ease of annotation.
In another embodiment, printouts of the images may be used and
manually segmented with pens/pencils, and the segmented printouts
may be scanned and recognized by the system using one or more
AI-algorithms (e.g., computer-vision based algorithms). Further, at
least a portion of such segmented images may be used as training
data that may be fed to the deep learning network at step 208, so a
GPU can learn from outlines of humans in the A-pose wearing any
clothes in any background. In one embodiment, the segmented images
from step 204 are utilized to train the segmentation DLN used in
step 106 of FIG. 1A.
[0120] In yet other embodiments, the segmentation training data may
have been automatically generated, for example as described in
related U.S. Ser. No. 16/517,391, filed on 19 Jul. 2019, cited
above.
[0121] At step 205, the annotator may then draw estimated
annotation (measurement) lines for each body feature under the
clothing using human intuition. As noted, accurate annotation lines
may be drawn by human annotators manually from the background. The
ability of human workers to determine the correct annotation lines
of photographed humans under any kind of clothing, especially by
skilled annotators who are experienced and who can provide accurate
and reliable body shape annotations, ensures high performance for
the system. The annotation lines may be drawn on any suitable
software platform, and may use a peripheral device (e.g., a smart
pen) for ease of annotation. In another embodiment, printouts of
the images may be used and manually annotated with pens/pencils,
and the annotated printouts may be scanned and recognized by the
system using one or more AI-algorithms (e.g., computer-vision based
algorithms). Further, at least a portion of such annotated images
may be used as training data that may be fed to the deep learning
network at step 210 below, so a GPU can learn from annotation lines
of humans in the A-pose wearing any clothes in any background.
[0122] In yet other embodiments, the annotation training data may
have been automatically generated, for example as described in
related U.S. Ser. No. 16/517,391, filed on 19 Jul. 2019, cited
above.
[0123] A starting point for any machine learning method such as
used by the deep learning component above is a documented dataset
containing multiple instances of system inputs and correct outcomes
(e.g., the training data). This data set can be used, using methods
known in the art, including but not limited to standardized machine
learning methods such as parametric classification methods,
non-parametric methods, decision tree learning, neural networks,
methods combining both inductive and analytic learning, and
modeling approaches such as regression models, to train the machine
learning system and to evaluate and optimize the performance of the
trained system. The quality of the output of the machine learning
system output depends on (a) the pattern parameterization, (b) the
learning machine design, and (c) the quality of the training
database. These components can be refined and optimized using
various methods. For example, the database can be refined by adding
datasets for new documented subjects. The quality of the database
can be improved, for example, by populating the database with cases
in which the customization was accomplished by one or more experts
in garment customization. Thus, the database will better represent
the expert's knowledge.
[0124] At step 206, actual human measurements for each body feature
(e.g., determined by a tailor, or 1D measurements taken from 3D
body scans) may be received to serve as ground-truth data. The
actual human measurements may be used as validation data and used
for training the algorithms used by the system. For example, the
actual human measurements may be used in minimizing an error
function or loss function (mean squared error, likelihood loss,
log-loss, hinge loss, etc.) associated with the machine learning
algorithms. In one embodiment, the annotation lines from step 205
and the ground-truth data from step 206 are utilized to train the
annotation DLN used in step 107 and the sizing ML step 108 of FIG.
1A.
[0125] In one embodiment, the human measurements may be received
from a user input (e.g., an input to a user device such as a
smartphone). In another embodiment, the human measurements may be
received from a network (e.g., the Internet), for example, through
a website. For example, a tailor may upload one or more
measurements to a website and the system may receive the
measurements. As noted, in another embodiment, the actual
measurements may be used to train and/or improve the accuracy of
the AI-based algorithmic results (e.g., deep learning models)
results, to be discussed below.
[0126] In yet other embodiments, the sizing ground truth training
data may have been automatically generated, for example as
described in related U.S. Ser. No. 16/517,391, filed on 19 Jul.
2019, cited above.
[0127] At step 208, the segmentation DLN may be trained on the
segmentation training data. In one embodiment, the segmentation DLN
may be trained using human body segmentation obtained in step 204.
For example, the segmentation DLN may be presented with labeled
data (e.g., an image of a user and associated actual body
segmentations) and may determine an error function (e.g., from a
loss function, as discussed above) based on the results of the
segmentation DLN and the actual body segmentation. The segmentation
DLN may be trained to reduce the magnitude of this error
function.
[0128] In another embodiment, the segmentation DLN may be validated
by accuracy estimation techniques like a holdout method, which may
split the data (e.g., all images including images having
corresponding segmentations, and images on which to extract
segmentations using the segmentation DLN and having no
corresponding segmentations) in a training and test set
(conventionally 2/3 training set and 1/3 test set designation) and
may evaluate the performance of the segmentation DLN model on the
test set. In another embodiment, a N-fold-cross-validation method
may be used, where the method randomly splits the data into k
subsets where k-1 instances of the data are used to train the
segmentation DLN model while the kth instance is used to test the
predictive ability of the segmentation DLN model. In addition to
the holdout and cross-validation methods, a bootstrap method may be
used, which samples n instances with replacement from the dataset,
can be used to assess the segmentation DLN model accuracy.
[0129] At step 210, one or more separate annotation DLNs for each
body feature may be trained, or alternatively a single annotation
DLN for the entire body may be trained. For example, sixteen
annotation DLNs, one for each of 16 different body features, may be
trained. In one embodiment, the annotation DLN may be trained using
the annotations obtained in step 205. For example, the annotation
DLN may be presented with labeled data (e.g., an image of a user
with line annotations) and may determine an error function (e.g.,
from a loss function, as discussed above) based on the results of
the annotation DLN and the actual annotations. The annotation DLN
may be trained to reduce the magnitude of this error function.
[0130] In another embodiment, an annotation DLN may be trained
specifically to draw annotation lines from a particular body
feature, for example, a specific body part, such as an arm, a leg,
a neck, and so on. In another embodiment, the training of the
annotation DLN for each body feature may be performed in series
(e.g., in a hierarchical manner, with groups of related body
features being trained one after the other) or in parallel. In
another embodiment, different training data sets may be used for
different annotation DLNs, the different annotation DLNs
corresponding to different body features or body parts. In one
embodiment, there may be more or less than sixteen DLNs for the
sixteen body parts, for example, depending on computational
resources. In another embodiment, the training of the annotation
DLNs may be performed at least partially in the cloud, to be
described below.
[0131] Also, at step 210, one or more sizing ML modules for each
body feature may be trained, or alternatively a single sizing ML
module for the entire body may be trained. In one embodiment, the
sizing ML module may be trained using the measurements obtained in
step 206. For example, the sizing ML module may be presented with
labeled data (e.g., an annotation line length and associated actual
measurement data) and may determine an error function (e.g., from a
loss function, as discussed above) based on the results of the
sizing ML module and the actual measurements. The sizing ML module
may be trained to reduce the magnitude of this error function.
[0132] In another embodiment, a sizing ML module may be trained
specifically to extract measurements from a particular body
feature, for example, a specific body part, such as an arm, a leg,
a neck, and so on. In another embodiment, the training of the
sizing ML module for each body feature may be performed in series
(e.g., in a hierarchical manner, with groups of related body
features being trained one after the other) or in parallel. In
another embodiment, different training data sets may be used for
different sizing ML modules, the sizing ML modules corresponding to
different body features or body parts. In one embodiment, there may
be more or less than sixteen sizing ML modules for the sixteen body
parts, for example, depending on computational resources. In
another embodiment, the training of the sizing ML modules may be
performed at least partially in the cloud, to be described
below.
[0133] At step 212, the trained segmentation DLN, annotation DLN,
and sizing ML module to be used in FIGS. 1A, 1B, and 1C may be
output. In particular, the segmentation DLN trained in step 208 is
output for use in step 105 in FIG. 1A. Similarly, the one or more
annotation DLNs trained in step 210 are output for use in step 107
in FIG. 1A. Finally, the sizing ML module trained in step 210 is
output for use in step 108 in FIG. 1A.
[0134] FIGS. 3-8 show illustrative diagrams of a graphical user
interface (GUI) corresponding to process steps in FIG. 2 that are
utilized to generate training data for training the segmentation
and annotation DLNs. FIG. 3 shows an illustrative diagram of a user
image showing a human body wearing clothes captured for training
the segmentation DLN. Although a specific user pose, the "A-pose,"
is shown in FIGS. 3-8, it will be understood to one of ordinary art
that any pose, such as the A-pose, hands on the side, or any other
pose is within the scope of the present invention. An optimal pose
would clearly show legs and arms separated from the body. One
advantage of the present invention is that a human can stand in
almost any reasonable pose, against any type of background. The
human does not need to stand against a blank background or make
special arrangements for where the photos are taken.
[0135] FIG. 4 shows an illustrative diagram of an annotator, or
operator, manually segmenting one or more features of the human
body under the clothing from the background for training the
segmentation DLN. In FIG. 4, the annotator is manually annotating
the location of the left leg under the clothing. Humans have lots
of experience of seeing other humans and estimating their body
shapes under the clothing, and this data is used for training the
segmentation DLN to perform a similar operation automatically on
new photos of unknown humans. FIG. 5 shows an illustrative diagram
of the body features of the human body segmented from the
background after all body features have been successfully annotated
by the annotator. This data is used to train the human segmentation
DLN. FIG. 5 provides the manually-annotated data that is used in
training the segmentation DLN in step 208 of FIG. 2. The
segmentation DLN trained in step 208 of FIG. 2, using the data
obtained in FIG. 5, is then used in step 105 in FIG. 1A.
[0136] FIG. 6 shows an illustrative diagram of the annotator
manually drawing annotation lines for training the annotation DLN.
This data is used to train the annotation DLN for automatically
drawing annotation lines for each body feature. FIG. 6 provides the
manually-annotated data that is used in training the annotation DLN
in step 210 of FIG. 2. The annotation DLN trained in step 210 of
FIG. 2, using the data obtained in FIG. 6, is then used in step 107
in FIG. 1A.
[0137] FIGS. 7-8 show illustrative diagrams of the annotation
process being repeated for the side views of the human. FIG. 7
shows another illustrative diagram (side view) of the annotator
manually segmenting one or more body features of the human body
under the clothing from the background for training the
segmentation DLN. FIG. 8 shows another illustrative diagram (side
view) of the annotator manually drawing annotation lines for
training the annotation DLN. Although only front and side views are
shown in FIGS. 3-8, one of ordinary skill in the art would
recognize that any other orientations of views, including 45 degree
views, top views, and so on, are within the scope of the present
invention, depending on the type of body measurement desired. For
example, a top photo of the top of a human head would be optimal
for head measurements used to manufacture custom hats. Similarly, a
front face only photo would be optimal for facial measurements used
for sizing glasses, optical instruments, and so forth. A close-up
photo of the front and back of a human hand can be used for sizing
custom gloves, custom PPE (personal protective equipment) for the
hands, custom nails, and so forth.
[0138] In one embodiment, it is also possible to deploy human
beings to assist the deep-learning networks in the calculation
process, analogous to directed supervised learning. A human
annotator may manually adjust, or edit, a result from the
segmentation DLN and/or the annotation DLNs to deliver even more
accurate sizing results. The "adjustment data" made by the human
annotator to the segmentation and annotation maps from the deep
learning networks can be used in a feedback loop back into the deep
learning networks to improve the DNL models automatically over
time.
Alternative Deep-Learning Network (DLN) Architectures
[0139] FIG. 9 shows an illustrative client-server diagram 900 for
implementing body measurement extraction, in accordance with one
embodiment of the invention. The client-side (user) 909 is shown at
the top, while the server-side 903 is shown at the bottom. The
client-side initiates the process by sending front and side images
at 902. After receiving the images, the server checks the images
for the correctness of the format and other formal checks at 904.
If images are not of the correct format or have other formal
problems at 905, such as wrong pose, poor contrast, too far or too
close, subject not in view, subject partially obstructed, and so
forth, the process returns to this information to the client at
901. At 901, an error message or other communication may be
displayed to the user, in one embodiment, to enable to user to
retake the images.
[0140] If the images are of the correct format and have no other
formal problems at 905, the images are pre-processed at 906 so that
they can be handled by the DLN (deep learning network). The images
are then processed through the DLN to determine sizing at 908, as
described in greater detail previously. The sizing results or
complete body measurements are returned at 910 from the server. The
client checks the sizing results at 912. If the sizing results have
any formal problems, for example being out-of-bounds, unreasonably
small or large, and so on, as determined at 913, the process
returns to 901, and similarly displays an error message or other
communication may be displayed to the user to enable to user to
retake the images. If the sizing results have no formal problems,
as determined at 913, the process ends with the complete body
measurements ready for use.
[0141] FIG. 10 shows a diagram of one example flow diagram 1000 for
body measurement determination (using separate segmentation DLN,
annotation DLN, and sizing ML module), in accordance with one
embodiment of the invention. In one embodiment, front and side
images are received from a user at 1002. The images are
pre-processed at 1004. As previously discussed, in some
embodiments, a pre-processing on the one or more images of the
user, such as a perspective correction, may be performed on the
front and side view photos, if needed. For example, the system may
use OpenCV, an open source machine vision library, and may make use
of features of the head in the front and side view photographs and
the user's height as references for perspective correction. Various
computer vision techniques may be utilized to further pre-process
the one or more images. Examples of pre-processing steps may
include, in addition to perspective correction, contrast, lighting,
and other image processing techniques to improve the quality of the
one or more images before further processing.
[0142] After pre-processing, the pre-processed images are sent to
the segmentation DLN at 1006 to generate the segmentation map, as
discussed previously. The segmentation map is aggregated with the
rest of the data at 1014. In parallel to the segmentation, in one
embodiment, the pre-processed images are also sent to annotation
DLN at 1008 to generate the annotation measurement lines, as
discussed previously. The annotation map is aggregated with the
rest of the data at 1014. The annotation map is provided, in one
embodiment, to sizing machine learning (ML) module 1010 to generate
the body feature measurements for each body feature that has been
segmented and annotated by measuring each annotation line, as
discussed previously. The sizing result is aggregated with the rest
of the data at 1014. The sizing result is output to one or more
external system(s) for various uses as described herein at 1012.
Finally, all of the aggregated and structured data, (1) the
pre-processed front and side images, (2) the segmentation map, (3)
the annotation map, and (4) the sizing result, that have been
aggregated at 1014 are stored in a database for further DLN
training at 1016.
[0143] FIG. 11 shows a diagram of another example flow diagram 1100
for body measurement determination (using a combined
segmentation-annotation DLN and sizing ML module), in accordance
with another embodiment of the invention. Front and side images are
received from a user at 1102, and the images are pre-processed at
1104, as previously discussed. Examples of pre-processing steps
include perspective correction, contrast, lighting, and other image
processing techniques to improve the quality of the one or more
images before further processing.
[0144] After pre-processing, the pre-processed images are sent
directly to the annotation DLN at 1018 to generate the annotation
map, as discussed previously. Instead of first performing body
feature segmentation, in this alternative embodiment, the
annotation lines are drawn directly on the images without
explicitly segmenting the body features from the background using a
specially-trained combined segmentation-annotation DLN that
effectively combines the features of both the segmentation DLN and
the annotation DLN (shown in the embodiment in FIG. 10) into a
single annotation DLN shown in FIG. 11. In effect, the body feature
segmentation is performed implicitly by the annotation DLN. The
annotation map is aggregated with the rest of the data at 1114.
[0145] The annotation map is provided, in one embodiment, to sizing
machine learning (ML) module 1110 to generate the body feature
measurements for each body feature that has been annotated by
measuring each annotation line, as discussed previously. The sizing
result is aggregated with the rest of the data at 1114. The sizing
result is output to one or more external system(s) for various uses
as described herein at 1112. Finally, all of the aggregated and
structured data, (1) the pre-processed front and side images, (2)
the annotation map, and (3) the sizing result, that have been
aggregated at 1114 are stored in a database for further DLN
training at 1116.
[0146] FIG. 12 shows a diagram of yet another example flow diagram
1200 for body measurement determination (using a combined sizing
DLN), in accordance with yet another embodiment of the invention.
Front and side images are received from a user at 1202, and the
images are pre-processed at 1204, as previously discussed. Examples
of pre-processing steps include perspective correction, contrast,
lighting, and other image processing techniques to improve the
quality of the one or more images before further processing.
[0147] After pre-processing, the pre-processed images are sent
directly to the sizing DLN at 1210 to generate the complete body
feature measurements, as discussed previously. Instead of first
performing body feature segmentation and annotation of the
measurement lines, followed by measurement of the lines, in this
alternative embodiment, the body feature are directly extracted
from the pre-processed images without explicitly segmenting the
body features from the background (and without explicitly drawing
the annotation lines) using a specially-trained sizing DLN that
effectively combines the features of the segmentation DLN, the
annotation DLN, and the measurement machine learning modules (shown
in the embodiment shown in FIG. 10) into a single sizing DLN shown
in FIG. 12. In effect, the body feature segmentation and the
annotation of the measurement lines is performed implicitly by the
sizing DLN.
[0148] The sizing result is aggregated with the rest of the data at
1214. The sizing result is output to one or more external system(s)
for various uses as described herein at 1212. Finally, all of the
aggregated and structured data, (1) the pre-processed front and
side images and (2) the sizing result, that has been aggregated at
1214 are stored in a database for further DLN training at 1216.
3D Model and Skeleton/Joint Position Model Embodiments
[0149] FIG. 13 shows a diagram of another example process flow 1300
for body measurement determination operations, in accordance with
example embodiments of the disclosure. At step 1302, user
parameters (e.g., height, weight, demographics, athleticism, and
the like) may be received from a user and/or receive parameters
auto-generated by a phone camera. In additional aspects, user
parameters may be determined automatically (e.g., using computer
vision algorithms or mined from one or more databases), or
determined from the user (e.g., user input). In another embodiment,
from these parameters, a body mass index (BMI) may be calculated.
As noted, the BMI (or any other parameter determined above) may be
used to calibrate the body weight to height. At step 1304, images
of the user (e.g., first and second images representing a full-body
front and a side view of the user) may be received, and an optional
third image may be received (for example, a 45-degree view between
the front and the side view, which may be used to enhance the
accuracy of the subsequent algorithms). In another embodiment, the
images may be obtained from the user device (e.g., mobile phone,
laptop, tablet, etc.). In another embodiment, the images may be
determined from a database (e.g., a social media database). In
another embodiment, the user may indicate whether he or she is
dressed in tight, normal, or loose clothing for more accurate
results.
[0150] At step 1306, a human segmentation (e.g., an extraction of
the human from a background of the images) may be performed, and a
3D model may be fitted against an extracted human. Moreover, a
three-dimensional shape may be estimated using a three-dimensional
modeling technique. In one embodiment, the system may utilize deep
learning techniques and/or OpenCV to extract the human body,
including clothing, from the background. Before performing this
step on data from a real user, the system may have been trained
first, for example, on sample photos of humans posing in different
environments in different clothing, with hands at 45 degrees
("A-pose").
[0151] At step 1308, a joint position and posture of the human may
be determined using skeleton detection; further, the determination
may be performed using a pose estimation algorithm such as
OpenPose, an open source algorithm, for pose detection. In one
embodiment, body pose estimation may include algorithms and systems
that recover the pose of an articulated body, which consists of
joints and rigid parts using image-based observations. In another
embodiment, OpenPose may include a real-time multi-person system to
jointly detect human body, hand, facial, and foot keypoints (in
total 135 keypoints) on single images. In one embodiment, a
keypoint may refer to a part of a person's pose that is estimated,
such as the nose, right ear, left knee, right foot, etc. The
keypoint contains both a position and a keypoint confidence score.
Further aspects of OpenPose functionality include, but not be
limited to, 2D real-time multi-person keypoint body estimation. The
functionality may further include the ability for the algorithms to
be run-time invariant to number of detected people. Another aspect
of its functionality may include, but may not be limited to, 3D
real-time single-person keypoint detection, including 3D
triangulation from multiple single views.
[0152] At step 1310, body sizing measurements may be determined
based on estimated three-dimensional shape, joint position and/or
posture. In another embodiment, the system may determine the sizes
of the body parts using inputs of height, weight, and/or other
parameters (e.g., BMI index, age, gender, etc.). In one embodiment,
the system may use, in part, a Virtuoso algorithm, an algorithm
providing standard DaVinci models of human body parts and relative
sizes of body parts.
[0153] Further, the system may generate and extract body
measurements by using an AI-based algorithm such as a DLN
algorithm, as described above, for example, by drawing measurement
lines from signals obtained from skeleton points. In particular,
the system may look at one or more skeleton points, calculate bone
angled against the edge of the user's body and draw measurement
lines in certain orientations or directions. Each measurement line
may be different for each body part and may be drawn differently.
The system may also use other algorithms, means, medians, and other
resources.
[0154] Further, the body measurements may be outputted to a user
device and/or a corresponding server. In one embodiment, the output
may be in the form of a text message, an email, a textual
description on a mobile application or web site, combinations
thereof, and the like.
[0155] At step 1312, the body sizing measurements may be updated to
reduce errors by using a supervised deep-learning algorithm that
makes use of training data, the training data comprising
manually-determined body detection underneath clothing. In some
aspects, any suitable deep learning architecture may be used such
as deep neural networks, deep belief networks and recurrent neural
networks, as described above. In another embodiment, training data
may be obtained from annotator input that extracts a human body
from a given photo, excluding clothing, from a background of the
photos, as described above. In short, the deep learning approaches
described above may be used, in some embodiments, in combination to
improve the accuracy and reliability of the 3D model and
skeleton/joint position approaches.
[0156] Accordingly, some embodiments of the present invention
include a computer-implemented method comprising receiving user
parameters from a user or a user device; receiving at least one
image from the user device, the at least one image including a
background and a human; performing human segmentation on the at
least one image to extract features associated with the human from
the background; performing skeleton detection on the extracted
features to determine joint positions of the human; performing body
detection based on the determined joint positions; and generating
body sizing measurements using the body detection.
[0157] Accordingly, other embodiments of the present invention
include a computer system, comprising a memory that stores
computer-executable components; a processor, operably coupled to
the memory, and that executes the computer-executable components
stored in the memory, wherein the computer-executable components
comprise a data collection component communicatively coupled with
the processor that receives user parameters from the user or a user
device and receives at least one image from the user device, the at
least one image including a background and a human; a data
extraction component communicatively coupled with the processor
that performs human segmentation on the at least one image to
extract features associated with the human from the background, and
that performs skeleton detection on the extracted features to
determine joint positions of the human; and a data analysis
component that performs body detection based on the determined
joint positions and generates the body sizing measurements using
the body detection.
Hardware, Software, and Cloud Implementation of the Present
Invention
[0158] As discussed, the data (e.g., photos, textual descriptions,
and the like) described throughout the disclosure can include data
that is stored on a database stored or hosted on a cloud computing
platform. It is to be understood that although this disclosure
includes a detailed description on cloud computing, below,
implementation of the teachings recited herein are not limited to a
cloud computing environment. Rather, embodiments of the present
invention are capable of being implemented in conjunction with any
other type of computing environment now known or later
developed.
[0159] Cloud computing can refer to a model of service delivery for
enabling convenient, on-demand network access to a shared pool of
configurable computing resources (e.g., networks, network
bandwidth, servers, processing, memory, storage, applications,
virtual machines, and services) that can be rapidly provisioned and
released with minimal management effort or interaction with a
provider of the service. This cloud model can include at least five
characteristics, at least three service models, and at least four
deployment models.
[0160] Characteristics may include one or more of the following.
On-demand self-service: a cloud consumer can unilaterally provision
computing capabilities, such as server time and network storage, as
needed automatically without requiring human interaction with the
service's provider. Broad network access: capabilities are
available over a network and accessed through standard mechanisms
that promote use by heterogeneous thin or thick client platforms
(e.g., mobile phones, laptops, and PDAs). Resource pooling: the
provider's computing resources are pooled to serve multiple
consumers using a multi-tenant model, with different physical and
virtual resources dynamically assigned and reassigned according to
demand. There is a sense of location independence in that the
consumer generally has no control or knowledge over the exact
location of the provided resources but can be able to specify
location at a higher level of abstraction (e.g., country, state, or
datacenter). Rapid elasticity: capabilities can be rapidly and
elastically provisioned, in some cases automatically, to quickly
scale out and rapidly released to quickly scale in. To the
consumer, the capabilities available for provisioning often appear
to be unlimited and can be purchased in any quantity at any time.
Measured service: cloud systems automatically control and optimize
resource use by leveraging a metering capability at some level of
abstraction appropriate to the type of service (e.g., storage,
processing, bandwidth, and active user accounts). Resource usage
can be monitored, controlled, and reported, providing transparency
for both the provider and consumer of the utilized service.
[0161] In another embodiment, Service Models may include the one or
more of the following. Software as a Service (SaaS): the capability
provided to the consumer is to use the provider's applications
running on a cloud infrastructure. The applications are accessible
from various client devices through a thin client interface such as
a web browser (e.g., web-based e-mail). The consumer does not
manage or control the underlying cloud infrastructure including
network, servers, operating systems, storage, or even individual
application capabilities, with the possible exception of limited
user-specific application configuration settings.
[0162] Platform as a Service (PaaS): the capability provided to the
consumer is to deploy onto the cloud infrastructure
consumer-created or acquired applications created using programming
languages and tools supported by the provider. The consumer does
not manage or control the underlying cloud infrastructure including
networks, servers, operating systems, or storage, but has control
over the deployed applications and possibly application hosting
environment configurations.
[0163] Infrastructure as a Service (IaaS): the capability provided
to the consumer is to provision processing, storage, networks, and
other fundamental computing resources where the consumer is able to
deploy and run arbitrary software, which can include operating
systems and applications. The consumer does not manage or control
the underlying cloud infrastructure but has control over operating
systems, storage, deployed applications, and possibly limited
control of select networking components (e.g., host firewalls).
[0164] Deployment Models may include one or more of the following.
Private cloud: the cloud infrastructure is operated solely for an
organization. It can be managed by the organization or a third
party and can exist on-premises or off-premises.
[0165] Community cloud: the cloud infrastructure is shared by
several organizations and supports a specific community that has
shared concerns (e.g., mission, security requirements, policy, and
compliance considerations). It can be managed by the organizations
or a third party and can exist on-premises or off-premises.
[0166] Public cloud: the cloud infrastructure is made available to
the general public or a large industry group and is owned by an
organization selling cloud services.
[0167] Hybrid cloud: the cloud infrastructure is a composition of
two or more clouds (private, community, or public) that remain
unique entities but are bound together by standardized or
proprietary technology that enables data and application
portability (e.g., cloud bursting for load-balancing between
clouds).
[0168] A cloud computing environment is service oriented with a
focus on statelessness, low coupling, modularity, and semantic
interoperability. At the heart of cloud computing is an
infrastructure that includes a network of interconnected nodes.
[0169] The cloud computing environment may include one or more
cloud computing nodes with which local computing devices used by
cloud consumers, such as, for example, personal digital assistant
(PDA) or cellular telephone, desktop computer, laptop computer,
and/or automobile computer system can communicate. Nodes can
communicate with one another. They can be group physically or
virtually, in one or more networks, such as Private, Community,
Public, or Hybrid clouds as described hereinabove, or a combination
thereof. This allows cloud computing environment to offer
infrastructure, platforms and/or software as services for which a
cloud consumer does not need to maintain resources on a local
computing device. It is understood that the types of computing
devices are intended to be exemplary only and that computing nodes
and cloud computing environment can communicate with any type of
computerized device over any type of network and/or network
addressable connection (e.g., using a web browser).
[0170] The present invention may be implemented using server-based
hardware and software. FIG. 14 shows an illustrative hardware
architecture diagram 1400 of a server for implementing one
embodiment of the present invention. Many components of the system,
for example, network interfaces etc., have not been shown, so as
not to obscure the present invention. However, one of ordinary
skill in the art would appreciate that the system necessarily
includes these components. A user-device is a hardware that
includes at least one processor 1440 coupled to a memory 1450. The
processor may represent one or more processors (e.g.,
microprocessors), and the memory may represent random access memory
(RAM) devices comprising a main storage of the hardware, as well as
any supplemental levels of memory e.g., cache memories,
non-volatile or back-up memories (e.g. programmable or flash
memories), read-only memories, etc. In addition, the memory may be
considered to include memory storage physically located elsewhere
in the hardware, e.g. any cache memory in the processor, as well as
any storage capacity used as a virtual memory, e.g., as stored on a
mass storage device.
[0171] The hardware of a user-device also typically receives a
number of inputs 1410 and outputs 1420 for communicating
information externally. For interface with a user, the hardware may
include one or more user input devices (e.g., a keyboard, a mouse,
a scanner, a microphone, a web camera, etc.) and a display (e.g., a
Liquid Crystal Display (LCD) panel). For additional storage, the
hardware my also include one or more mass storage devices 1490,
e.g., a floppy or other removable disk drive, a hard disk drive, a
Direct Access Storage Device (DASD), an optical drive (e.g. a
Compact Disk (CD) drive, a Digital Versatile Disk (DVD) drive,
etc.) and/or a tape drive, among others. Furthermore, the hardware
may include an interface one or more external SQL databases 1430,
as well as one or more networks 1480 (e.g., a local area network
(LAN), a wide area network (WAN), a wireless network, and/or the
Internet among others) to permit the communication of information
with other computers coupled to the networks. It should be
appreciated that the hardware typically includes suitable analog
and/or digital interfaces to communicate with each other.
[0172] The hardware operates under the control of an operating
system 1470, and executes various computer software applications
1460, components, programs, codes, libraries, objects, modules,
etc. indicated collectively by reference numerals to perform the
methods, processes, and techniques described above.
[0173] The present invention may be implemented in a client server
environment. FIG. 15 shows an illustrative system architecture 1500
for implementing one embodiment of the present invention in a
client server environment. User devices 1510 on the client side may
include smart phones 1512, laptops 1514, desktop PCs 1516, tablets
1518, or other devices. Such user devices 1510 access the service
of the system server 1830 through some network connection 1520,
such as the Internet.
[0174] In some embodiments of the present invention, the entire
system can be implemented and offered to the end-users and
operators over the Internet, in a so-called cloud implementation.
No local installation of software or hardware would be needed, and
the end-users and operators would be allowed access to the systems
of the present invention directly over the Internet, using either a
web browser or similar software on a client, which client could be
a desktop, laptop, mobile device, and so on. This eliminates any
need for custom software installation on the client side and
increases the flexibility of delivery of the service
(software-as-a-service), and increases user satisfaction and ease
of use. Various business models, revenue models, and delivery
mechanisms for the present invention are envisioned, and are all to
be considered within the scope of the present invention.
[0175] In general, the method executed to implement the embodiments
of the invention, may be implemented as part of an operating system
or a specific application, component, program, object, module or
sequence of instructions referred to as "computer program(s)" or
"computer code(s)." The computer programs typically comprise one or
more instructions set at various times in various memory and
storage devices in a computer, and that, when read and executed by
one or more processors in a computer, cause the computer to perform
operations necessary to execute elements involving the various
aspects of the invention. Moreover, while the invention has been
described in the context of fully functioning computers and
computer systems, those skilled in the art will appreciate that the
various embodiments of the invention are capable of being
distributed as a program product in a variety of forms, and that
the invention applies equally regardless of the particular type of
machine or computer-readable media used to actually effect the
distribution. Examples of computer-readable media include but are
not limited to recordable type media such as volatile and
non-volatile memory devices, floppy and other removable disks, hard
disk drives, optical disks (e.g., Compact Disk Read-Only Memory (CD
ROMS), Digital Versatile Disks, (DVDs), etc.), and digital and
analog communication media.
Example Use Cases of the Present Invention
[0176] FIG. 16 is an illustrative diagram of a use case of the
present invention in which a single camera on a mobile device is
used to capture human body measurements, showing a front view of a
human in typical clothing standing against a normal background. In
some embodiments, the mobile device comprises at least one camera,
a processor, a non-transitory storage medium, and a communication
link to a server. In one embodiment, the one or more photos of the
user's body are transmitted to a server that performs the
operations described herein. In one embodiment, the one or more
photos of the user's body are analyzed locally by the processor of
the mobile device. The operations performed return one or more body
measurements, which may be stored on the server, as well as
presented to the user. In addition, the body measurements may then
be utilized for many purposes, including but not limited to,
offering for sale to the user one or more custom garments, custom
glasses, custom gloves, custom body suites, custom PPE (personal
protection equipment), custom hats, custom diet regiments, custom
exercise, gym, and workout routines, and so on. Without loss of
generality, the body measurements may be output, transmitted,
and/or utilized for any purpose for which body measurements are
useful.
[0177] Finally, FIGS. 17-23 show illustrative mobile graphical user
interfaces (GUIs) in which some embodiments of the present
invention have been implemented. FIG. 17 shows an illustrative
diagram of the mobile device GUI showing user instructions for
capturing a front view photo, according to one embodiment of the
present invention. FIG. 18 shows an illustrative diagram of the
mobile device GUI requesting the user to enter their height (and
optionally other demographic information, such as weight, age,
etc.) and to select their preferred fit style (tight, regular, or
loose fit.), according to one embodiment of the present invention.
FIG. 19 shows an illustrative diagram of the mobile device GUI for
capturing the front view photo, according to one embodiment of the
present invention. FIG. 20 shows another illustrative diagram of
the mobile device GUI for capturing the front view photo with an
illustrative A-pose shown in dotted lines, according to one
embodiment of the present invention. FIG. 21 shows an illustrative
diagram of the mobile device GUI for capturing the side view photo,
according to one embodiment of the present invention. FIG. 22 shows
an illustrative diagram of the mobile device GUI that is displayed
while the system processes the photos to extract the body
measurements, according to one embodiment of the present invention.
Lastly, FIG. 23 shows an illustrative diagram of the mobile device
GUI showing a notification screen when the body measurements have
been successfully extracted, according to one embodiment of the
present invention.
[0178] The present invention has been successfully implemented
resulting in sub 1 cm accuracy body measurements relative to a
human tailor. The system is able to use just two photos and achieve
accuracy comparable to a human tailor. The system does not require
the use of any specialized hardware sensors, does not require the
user to stand against any special background, does not require
special lighting, can be used with photos taken at any distance,
and with the user wearing any type of clothing. The result is a
body measurement system that works with any mobile device so that
anyone can easily take photos of themselves and benefit from
automatic full body measurement extraction.
[0179] One of ordinary skill in the art knows that the use cases,
structures, schematics, and flow diagrams may be performed in other
orders or combinations, but the inventive concept of the present
invention remains without departing from the broader scope of the
invention. Every embodiment may be unique, and methods/steps may be
either shortened or lengthened, overlapped with the other
activities, postponed, delayed, and continued after a time gap,
such that every user is accommodated to practice the methods of the
present invention.
[0180] Although the present invention has been described with
reference to specific exemplary embodiments, it will be evident
that the various modification and changes can be made to these
embodiments without departing from the broader scope of the
invention. Accordingly, the specification and drawings are to be
regarded in an illustrative sense rather than in a restrictive
sense. It will also be apparent to the skilled artisan that the
embodiments described above are specific examples of a single
broader invention which may have greater scope than any of the
singular descriptions taught. There may be many alterations made in
the descriptions without departing from the scope of the present
invention.
* * * * *