Abstract: Masked Autoencoder (MAE) has recently been shown to be effective in pre-training Vision Transformers (ViT) for natural image analysis. By reconstructing full images from partially masked ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results