normalclone/fashion-ai-analysis

Extract attribute of clothes detected in image.

FashionAI-Analysis

This is a project for extracting attribute of clothes detected in image.

Dependencies

• python >= 3.6
• pytorch >= 1.2
• opencv
• matplotlib
• RoiAlign

Installation

1. Download & install cuda 10.2 toolkit here
2. Download & install anaconda python 3.7 version
3. Install requirements

How to test the model

1. Download yolo-v3 model from here and put in 'root directory'.

2. Downoad model from here and put in 'root directory'.

3. Run file_demo.py to run each image file demostration

4. Run cam_demo.py to run web-cam demostration (not tested)

Clothing multi-attributes definition

1

2

3

4

5

6

7

8

9

10

11

12

13

GT values

Top color(14)

Top pattern(6)

Top gender(2)

Top season(4)

Top type(7)

Top sleeves(3)

Bottom color(14)

Bottom pattern(6)

Bottom gender(2)

Bottom season(4)

Bottom length(2)

Bottom type(2)

leg pose(3)

0

null

null

null

null

null

null

null

null

null

null

null

null

null

1

white

plain

man

spring

shirt

short sleeves

white

plain

man

spring

short pants

pants

standing

2

black

checker

woman

summer

jumper

long sleeves

black

checker

woman

summer

long pants

skirt

sitting

3

gray

dotted

autunm

jacket

no sleeves

gray

dotted

autunm

lying

4

pink

floral

winter

vest

pink

floral

winter

5

red

striped

parka

red

striped

6

green

mixed

coat

green

mixed

7

blue

dress

blue

8

brown

brown

9

navy

navy

10

beige

beige

11

yellow

yellow

12

purple

purple

13

orange

orange

14

mixed

mixed

A nice example

fashion-ai-analysis's Issues

Roi problem on mac

Hello,
I am using this library on windows and it works for me, but when I want to run the library on Mac, the Roi library becomes a problem and after installing the suitable compiler, it gives me the error " module pytorch is not found" while I have it in my libraries. Can I replace the syntax with other available libraries except ROI?

support by yolov3-tiny

Your code is great.

yolov3 is slow.
I want to use yolov3-tiny.

The cfg and weights files were modified.
I didn't get yolov3.get_feature(). (None)

What is the way to use tiny?

Your model is not available anymore

No yolo model on drive.

encoder-12-1170.ckpt

encoder-12-1170.ckpt How can we get this?

Queries

@LittleWat @normalclone THanks for sharing the code base its really helpful , i have few queries

1. can you share the training pipeline code
2. what is the inference time taken for single image on ur system
3. Can we train with custom attributes if so how should we provide
4. What is the dataset used for ur training

Thanks in advance

Crashes for specific images (batch_index 2 out of range [0, 2).)

I'm running this on hundereds of images that are similar in size and resolution, yet it crashes giving this error Error: batch_index 2 out of range [0, 2).
Any preprocessing I can do to the images so it works?
Thanks.