This is the competition of DeepFashion2 Challenge 2019 - Track 1 Clothes Landmark Estimation, more details can be found here
Details of the training data
Place the data as the following structure:
datasets
└── fashion
├── train
│ ├── annos
│ ├── deepfashion2_coco.json
│ └── image
├── train.zip
├── validation
│ ├── annos
│ ├── deepfashion2_coco.json
│ └── image
└── validation.zip
-
Transform the deepfashion2 dataset to coco-style
python pre-task/fashion2coco.py
This would generate two annotation files of coco-style, which locate at
data/train/deepfashion2_coco.json
anddata/validation/deepfashion2_coco.json