This project uses annotated images of soles from the shoe to train the model with UNET architecture in order to obtain a mask with predicted sole outline and location.This approach can be used in the shoe recycling industry to locate the sole and separate it from the upper part of the shoe, which would be an alternative method of sepparating the rubber part of the show within shoe-recycling industry.
Implemented in PyTorch.
Used dataset : UT Zappos50K - https://vision.cs.utexas.edu/projects/finegrained/utzap50k/
