MaskGIT: Masked Generative Image Transformer

Official Jax Implementation of the CVPR 2022 Paper

Summary

MaskGIT is a novel image synthesis paradigm using a bidirectional transformer decoder. During training, MaskGIT learns to predict randomly masked tokens by attending to tokens in all directions. At inference time, the model begins with generating all tokens of an image simultaneously, and then refines the image iteratively conditioned on the previous generation.

Running pretrained models

Class conditional Image Genration models:

Dataset	Resolution	Model	Link	FID
ImageNet	256 x 256	Tokenizer	checkpoint	2.28 (reconstruction)
ImageNet	512 x 512	Tokenizer	checkpoint	1.97 (reconstruction)
ImageNet	256 x 256	MaskGIT Transformer	checkpoint	6.06 (generation)
ImageNet	512 x 512	MaskGIT Transformer	checkpoint	7.32 (generation)

You can run these models for class-conditional image generation and editing in the demo Colab.

Training

[Coming Soon]

BibTeX

@InProceedings{chang2022maskgit,
  title = {MaskGIT: Masked Generative Image Transformer},
  author={Huiwen Chang and Han Zhang and Lu Jiang and Ce Liu and William T. Freeman},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2022}
}

Disclaimer