iMat_Fashion_2018
├── data
│ ├── train.json
│ ├── train.csv
│ ├── validation.json
│ ├── val.csv
│ ├── train
│ │ ├── 1.jpg
│ │ ├── 2.jpg
│ │ └── ...
│ └── val
│ ├── 1.jpg
│ ├── 2.jpg
│ └── ...
├── download_image.ipynb
├── label_map_228.csv
├── README.md
├── Train.ipynb
└── utils
├── dataset.py
├── download_utils.py
└── train.py
- Cause of file size problem, I only provide source codes not the data
- If you want to follow up, see Section 2. Steps to Follow-up
- Download train.json and place it following Section 1. Directory Structure [link]
- Download validation.json and place it following Section 1. Directory Structure [link]
- Run download_image.ipynb
- this scripts make train.csv, val.csv and download train images, val images using multiprocessing
- train.csv, val.csv : labels of each train/val images
- Run Train.ipynb
- Baseline is provided using resnet of pytorch with 18, 34, 50, 101 depth
- Official dataset github : imat_fashion_comp