This model is the VQGAN+ tokenizer with a vocabulary size of 12 bits. It uses a downsampling factor of 16 and is trained on ImageNet for images of resolution 256.

You can find more details on the project page and in the paper.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train markweber/vqgan_plus_12bit

Paper for markweber/vqgan_plus_12bit

MaskBit: Embedding-free Image Generation via Bit Tokens

Paper • 2409.16211 • Published Sep 24, 2024 • 17

Evaluation results

rFID on ILSVRC/imagenet-1k
self-reported

1.390
InceptionScore on ILSVRC/imagenet-1k
self-reported

193.900
LPIPS on ILSVRC/imagenet-1k
self-reported

0.315
PSNR on ILSVRC/imagenet-1k
self-reported

21.000
SSIM on ILSVRC/imagenet-1k
self-reported

0.550
CodebookUsage on ILSVRC/imagenet-1k
self-reported

1.000