My euro coin dataset for image classification experiments. Front only.
coin-dataset
├── original
│ ├── 10c [1025 images]
│ ├── 1c [1150 images]
│ ├── 1e [1027 images]
│ ├── 20c [1058 images]
│ ├── 2c [1005 images]
│ ├── 2e [1041 images]
│ ├── 50c [1031 images]
│ └── 5c [1088 images]
├── raw
│ ├── raw1
│ │ ├── 10c_8
│ │ │ ├── IMG_20190707_010612.jpg
│ │ │ ├── IMG_20190707_010616.jpg
│ ...
├── extract_coins.py
├── modify_dataset.py
├── README.md
└── split_dataset.py
With extract_coins.py, pictures found in raw folder (not uploaded) are analyzed, and coins extracted. No elaboration is done on the images.
With modify_dataset.py the dataset is elaborated and standardized, there are several options:
- equalize the images using CLAHE on L channel (Lab)
- mask the background
- resize the images
Run modify_dataset.py -h for a full list of options.
The dataset can be split in train/validation/test sets using split_dataset.py.