XPixelGroup/HAT
A Hybrid Attention Transformer architecture for image super-resolution and restoration, achieving state-of-the-art results on benchmark datasets.
Not currently ranked — collecting fresh signals.
star history
HAT introduces a hybrid attention mechanism to activate more pixels in image super-resolution transformers, improving detail reconstruction. The method combines channel attention and spatial attention to better model long-range dependencies in high-resolution image generation. Published at CVPR 2023 with extended TPAMI version, it demonstrates superior performance across standard benchmarks including Set5, Set14, Urban100, and Manga109.
Frequently asked
- What is XPixelGroup/HAT?
- A Hybrid Attention Transformer architecture for image super-resolution and restoration, achieving state-of-the-art results on benchmark datasets.
- Is HAT open source?
- Yes — XPixelGroup/HAT is open source, released under the Apache-2.0 license.
- What language is HAT written in?
- XPixelGroup/HAT is primarily written in Python.
- How popular is HAT?
- XPixelGroup/HAT has 1.6k stars on GitHub.
- Where can I find HAT?
- XPixelGroup/HAT is on GitHub at https://github.com/XPixelGroup/HAT.