Summary of Key-graph Transformer For Image Restoration, by Bin Ren et al.
Key-Graph Transformer for Image Restoration
by Bin Ren, Yawei Li, Jingyun Liang, Rakesh Ranjan, Mengyuan Liu, Rita Cucchiara, Luc Van Gool, Nicu Sebe
First submitted to arxiv on: 4 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG); Image and Video Processing (eess.IV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Image restoration (IR) is a crucial task in computer vision that requires effective integration of global information. However, current transformer-based methods struggle with high input resolutions due to computational expenses. Moreover, the self-attention mechanism can be prone to considering irrelevant cues from unrelated objects or regions, leading to inefficiencies. To address these challenges, we propose the Key-Graph Transformer (KGT), which views patch features as graph nodes and uses a sparse yet representative Key-Graph constructor. This approach enables efficient attention mechanisms within each window with linear computational complexity. Our experiments across 6 IR tasks demonstrate state-of-the-art performance and advancements both quantitatively and qualitatively. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Image restoration is important for computer vision, but current methods struggle to use global information because it’s too computationally expensive. This can also lead to using information that isn’t helpful. To solve this problem, we created a new type of transformer called the Key-Graph Transformer (KGT). It looks at patch features like graph nodes and uses a special way to connect them. This makes attention mechanisms more efficient and accurate. We tested KGT on 6 different image restoration tasks and it performed better than other methods. |
Keywords
* Artificial intelligence * Attention * Self attention * Transformer
