- based on tensorflow and keras
- use DeepFashion2 dataset to train Mask R-CNN
- origin code from https://github.com/Manishsinghrajput98/Deepfashion2_Training
- download DeepFashion2 dateset in here and coco pre-trained model in here
- git clone this repo to local (git clone https://github.com/Xlgd/Mask_R-CNN_DeepFashion2)
- install anaconda and create a new environment (conda create -n env_name python=3.6)
- activate the environment (conda activate env_name)
- install packages with requirement.txt (while read requirement; do conda install --yes $requirement || pip install $requirement; done < requirement.txt)
- convert DeepFashin2 json to coco format json (use tools/deepfashion2coco.py)
- add image path and coco format json path in main.py (line 44 ~ 47)
- begin training (python main.py train --weights=coco)