Pre-processing to convert imagenet-1k to PyTorch's ImageFolder arrangement

https://huggingface.co/datasets/ILSVRC/imagenet-1k

I modified it a bit to work with the following code. Thanks so much for the following code👍

https://stackoverflow.com/a/78633633

Setup

git clone https://github.com/saten-private/imagenet-1k-image-folder-preprocessing.git

Procedure

train

1. Create a train folder under the repository.
2. Download train_images_X.tar.gz from here.
3. Extract the images from train_images_X.tar.gz into the train folder.
4. Run python train_preprocessing.py. The process divides the train folder into folders with class names and relocates them.

val

1. Create a val folder under the repository.
2. Download val_images.tar.gz from here.
3. Extract the images from val_images.tar.gz into the val folder.
4. Run python val_preprocessing.py. The process divides the val folder into folders with class names and relocates them.