Research map / Video / Video segmentation and tracking
Mining Relations among Cross-Frame Affinities for Video Semantic Segmentation
Guolei Sun, Yun Liu, Hao Tang, Ajad Chhatkuli, Le Zhang, Luc Van Gool
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 more reference frames (3 refs at -9,-6,-3 vs single ref at -3) (ablation) | MRCFA MiT-B1 · VSPW val | mIoU: 38.1 -> 38.9 | Improved |
| add more reference frames (3 refs vs single ref at -6) (ablation) | MRCFA MiT-B1 · VSPW val | mVC8: 85.1 -> 88.8 | Improved |
| add a single reference frame to SegFormer baseline (-3) (ablation) | SegFormer MiT-B1 · VSPW val | mIoU: 36.5 -> 38.1 | Improved |
| keep all tokens (p=100%) instead of selecting 50% (ablation) | MRCFA MiT-B1 · VSPW val | mIoU: 38.9 -> 39.4 mIoU; memory 903 -> 1068, FPS 40.1 -> 32.9 | Mixed |
| reduce token selection ratio to 10% (ablation) | MRCFA MiT-B1 · VSPW val | mIoU: 38.9 -> 35.9 | Hurt |
| use three feature scales instead of last stage only (ablation) | MRCFA MiT-B1 · VSPW val | mIoU: 37.5 -> 38.9 | Improved |
| use three feature scales instead of two (ablation) | MRCFA MiT-B1 · VSPW val | mIoU: 38.1 -> 38.9 | Improved |
| add MAA to SAR in the affinity decoder (ablation) | Affinity Decoder MiT-B1 · VSPW val | mIoU: 37.8 -> 38.9 | Improved |
| add SAR to MAA in the affinity decoder (ablation) | Affinity Decoder MiT-B1 · VSPW val | mIoU: 37.4 -> 38.9 | Improved |
| replace Feature Pyramid with affinity decoder (ablation) | MRCFA MiT-B1 · VSPW val | mIoU: 37.8 -> 38.9 | Improved |
| replace SegFormer with MRCFA | SegFormer MiT-B5 · VSPW val | mIoU: 48.2 -> 49.9 | Improved |
| replace SegFormer with MRCFA | SegFormer MiT-B1 · Cityscapes val | mIoU: 74.1 -> 75.1 | Improved |
Closest papers by meaning
- Learning Local and Global Temporal Contexts for Video Semantic Segmentation (TPAMI 2024 · similarity 0.88)
- Coarse-To-Fine Feature Mining for Video Semantic Segmentation (CVPR 2022 · similarity 0.88)
- Exploring Temporal Feature Correlation for Efficient and Stable Video Semantic Segmentation (AAAI 2024 · similarity 0.87)
- Learning Spatial-Semantic Features for Robust Video Object Segmentation (ICLR 2025 · similarity 0.87)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · similarity 0.86)
- Attention (as Discrete-Time Markov) Chains (NeurIPS 2025 · similarity 0.86)
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.
