Research map / Architectures and efficiency / Attention and transformers
HRFormer: High-Resolution Vision Transformer for Dense Predict
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add 3x3 depth-wise conv to FFN (ablation) | HRFormer-T · COCO pose val | AP: 66.88 -> 70.92 | Helped |
| add 3x3 depth-wise conv to FFN (ablation) | HRFormer-T · PASCAL-Context | mIoU: 46.84 -> 49.74 | Helped |
| add 3x3 depth-wise conv to FFN (ablation) | HRFormer-T · ImageNet-1K | top-1 acc: 77.83 -> 78.48 | Helped |
| add 3x3 depth-wise conv to FFN (ablation) | Swin-T · ImageNet-1K | top-1 acc: 81.3 -> 82.2 | Helped |
| add 3x3 depth-wise conv to FFN (ablation) | IntraWin-T · ImageNet-1K | top-1 acc: 80.2 -> 82.3 | Helped |
| DW conv in FFN instead of shifted windows (ablation) | HRFormer-T · PASCAL-Context | mIoU: 43.3 -> 49.7 | Helped |
| DW conv in FFN instead of shifted windows (ablation) | HRFormer-T · COCO pose val | AP: 67.3 -> 70.9 | Helped |
| replace conv basic blocks with transformer blocks (ablation) | HRNet-T · ImageNet-1K | top-1 acc: 76.5 -> 78.5 | Helped |
| replace conv basic blocks with transformer blocks (ablation) | HRNet-B · PASCAL-Context | mIoU: 55.2 -> 58.5 | Helped |
| HRFormer-B vs HRNet-W48 backbone | pose estimator · COCO pose val | AP: 76.3 -> 77.2 | Helped |
| HRFormer-B vs SETR-MLA | segmentation · ADE20K val | mIoU: 50.3 vs 50.0 | No clear effect |
| HRFormer-B vs Swin-B | classification · ImageNet-1K | top-1 acc: 83.5 vs 82.8 | Hurt |
Closest papers by meaning
- Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation (CVPR 2022 · similarity 0.88)
- Win-Win: Training High-Resolution Vision Transformers from Two Windows (ICLR 2024 · similarity 0.88)
- SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation (AAAI 2024 · similarity 0.86)
- CatFormer: Category-Level 6D Object Pose Estimation with Transformer (AAAI 2024 · similarity 0.85)
- LaVin-DiT: Large Vision Diffusion Transformer (CVPR 2025 · similarity 0.85)
- SDPose: Tokenized Pose Estimation via Circulation-Guide Self-Distillation (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.