README.md

Fashion-AI-segmentation using Deep Learning

A New Approach by using the Blend of Image-Processing Technique and Deep-Learning Algorithm to Segment any Fashion and e-commerce Retail Images. The code can be used for any industry on any images and the core algortithm is 'grab-cut algorithm" with the blend of Deep-Learning Convolutional Neural Networks. The Repo is designed in a preview way and its limited for fashion Images with auto-segmenting Top-wear clothes(Example: Tshirt, shirts) and Full-body clothes(salwar,gowns, shirt-pants-shoes)*

Image-Processing Resource

https://en.wikipedia.org/wiki/GrabCut

Deep Learning

https://en.wikipedia.org/wiki/Deep_learning

Package Requirements

1.Python

2.OpenCV 3.1.0

3.Keras with tensorflow backend

4.Pandas

5.NumPy

Demo (AI-Segmentaion)

Note: Demo Annotation shouldn't be replaced, adding new will not enable the code adaptation to new classes of images.(The demo phase classes : Fashion full-body, Top-wear)

1.clone the Repo to your local pc ensuring that all the package requirements satisfied.

2.Run the code from the terminal python fashion.py image1.jpg /Users/demo/save

3.argument1 -- image_name -- image1.jpg, argument2 -- save_directory -- /Users/demo/

4.Visualize the results in your save_directory.

Demo (Instance Segmentation)

instance segmentation image

Some Results

To run the project:

python instance_segmentation.py image_name save_dir

Author

Anish Josh

josh.anish1@gmail.com