Closed LightingMc closed 11 months ago
It's the default train/val splits for imagenet.
Where can I find details of that?
It already comes divided into train and val. You can download it from multiple sources, e.g. https://huggingface.co/datasets/imagenet-1k/tree/main
What is the train-test split used for the pre-training on ImageNet? Do you have a file that describes which image goes where? What is the split percentage?