ImageNet 64x64

The 64x64 ImageNet dataset is used for training a vector-quantized variational auto-encoder, encoding images into a tensor of latents.

BibTex: