fashion-parser

Fashion segmentation based on Matterport's mask r-cnn implementation using imaterialist-fashion-2019-FGVC6 data set.

Usage

• Edit commons/config.py for workspace details, folder paths etc.
• Hyper parameters are set in commons/fashion_config.py which is inherited from Matterport's config
• Use experiment.py to train and create a submission file;

python experiment.py --epochs <number of epochs> --val_split <between 0 - 1, ratio of validation samples>

Example Results

TODO

• Switch to Keras
• Change to Mask R-CNN
• Update Readme
• Serve with Flask