Research map / Detection and segmentation / CNN features
Context-Guided Spatial Feature Reconstruction for Efficient Semantic Segmentation
Zhenliang Ni, Xinghao Chen, Yingjie Zhai, Yehui Tang, Yunhe Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 Rectangular Self-Calibration Module for pyramid context extraction (ablation) | CGRSeg-T · ADE20K val | mIoU (%): 41.34 -> 42.57 | Improved |
| add RCM for spatial feature reconstruction (ablation) | CGRSeg-T · ADE20K val | mIoU (%): 42.57 -> 43.60 | Improved |
| add Dynamic Prototype Guided head (ablation) | CGRSeg-T · ADE20K val | mIoU (%): 40.86 -> 41.34 | Improved |
| use DPG head instead of HamHead (ablation) | CGRSeg-T · ADE20K val | mIoU (%): 42.1 -> 43.6 | Improved |
| use rectangular self-calibration attention instead of self-attention (ablation) | CGRSeg-T · ADE20K val | mIoU (%): 39.9 -> 43.6 | Improved |
| model key area with broadcast addition instead of multiplication (ablation) | RCA · ADE20K val | mIoU (%): 43.08 -> 43.61 | Improved |
| add shape self-calibration strip convolutions (1x11, 11x1) (ablation) | RCA · ADE20K val | mIoU (%): 41.27 -> 43.61 | Improved |
| use 3x3 depthwise conv in fusion function instead of 1x1 (ablation) | RCA · ADE20K val | mIoU (%): 41.70 -> 43.39 | Improved |
| use 5x5 local conv instead of 3x3 in fusion (ablation) | RCA · ADE20K val | mIoU (%): 43.39 -> 42.99 | Hurt |
| swap EfficientFormerV2 backbone for MSCAN-T | CGRSeg-T · ADE20K val | mIoU (%): SegNeXt 41.1 -> CGRSeg 42.6 | Improved |
Closest papers by meaning
- SegMAN: Omni-scale Context Modeling with State Space Models and Local Attention for Semantic Segmentation (CVPR 2025 · similarity 0.85)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.84)
- APSeg: Auto-Prompt Network for Cross-Domain Few-Shot Semantic Segmentation (CVPR 2024 · similarity 0.84)
- MaSS13K: A Matting-level Semantic Segmentation Benchmark (CVPR 2025 · similarity 0.83)
- SegGraph: Leveraging Graphs of SAM Segments for Few-Shot 3D Part Segmentation (NeurIPS 2025 · similarity 0.83)
- RG-SAN: Rule-Guided Spatial Awareness Network for End-to-End 3D Referring Expression Segmentation (NeurIPS 2024 · similarity 0.83)
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.
