HRNet/HRFormer
A Vision Transformer architecture that maintains high-resolution representations for dense prediction tasks like pose estimation and semantic segmentation.
Not currently ranked — collecting fresh signals.
star history
HRFormer adapts the transformer architecture for dense computer vision tasks by using multi-resolution parallel design inspired by HRNet, combined with local-window self-attention for computational efficiency. The model performs human pose estimation and semantic segmentation by processing image features at multiple resolutions simultaneously, addressing the memory and computation limitations of standard Vision Transformers that produce only low-resolution outputs.
Frequently asked
- What is HRNet/HRFormer?
- A Vision Transformer architecture that maintains high-resolution representations for dense prediction tasks like pose estimation and semantic segmentation.
- Is HRFormer open source?
- Yes — HRNet/HRFormer is open source, released under the MIT license.
- What language is HRFormer written in?
- HRNet/HRFormer is primarily written in Python.
- How popular is HRFormer?
- HRNet/HRFormer has 524 stars on GitHub.
- Where can I find HRFormer?
- HRNet/HRFormer is on GitHub at https://github.com/HRNet/HRFormer.