Download Imagenet weights here.
Requirements:
- Python 3.7
- Tensorflow 1.15
Usage:
- training:
python main.py
- predict:
python predict.py
Notes:
- ./files_npy/ : folder contatining the following npy files
- train_files.npy
- trains_labels.npy
- val_files.npy
- val_labels.npy
- test_files.npy
- test_labels.npy
*_files.npy contains filenames of images randomly chosen for each set.
*_labels.npy contains integer labels for the images setted in *_files.npy (must be in the same order).
- ./data/ contains the images.
- ./tmp/ where the weights are saved during training.
- ./losses/ where the loss and accuracy metrics are saved during training.