Add Restormer

[tushdon2]

Model description

Restormer: Efficient Transformer for High-Resolution Image Restoration was published in CVPR 2022, which introduced a new Vision Transformer based architecture for Image Restoration tasks like Deraining, Motion Deblurring, Defocus Deblurring and Denoising. It reduced the time complexity of Self Attention in Vision Transformers from O(n²) to O(n) by introducing Multi-Dconv Head Transposed Attention. It also introduced Gated-Dconv Feed-Forward Network.

@manyana72 and I would like to add this model to Huggingface.

cc: @NielsRogge

Open source status

• [X] The model implementation is available
• [X] The model weights are available

Provide useful links for the implementation

Paper, Code Implementation and pretrained model weights