CVPR'22 paper "Restormer: Efficient Transformer for High-Resolution Image Restoration"
Created on March 24|Last edited on April 15
Comment
Recently, Transformers have shown significant performance gains on natural language and high-level vision tasks. However, since their computational complexity grows quadratically with the spatial resolution, they aren’t feasible to apply to most restoration tasks involving high-resolution images.
The authors propose an efficient Transformer model by making several key designs in the building blocks to capture long-range pixel interactions, while still remaining applicable to large images. The model is named Restoration Transformer (Restormer) and it achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising).
Check out the code and the full paper below:
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.