Research map / Video / Video segmentation and tracking
Learning Local and Global Temporal Contexts for Video Semantic Segmentation
Guolei Sun, Yun Liu, Henghui Ding, Min Wu, 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 static contexts to SegFormer (ablation) | SegFormer MiT-B1 · VSPW val (ablation) | mIoU: 36.5 -> 37.7 | Improved |
| add motional contexts on top of static contexts (full CFFM) (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 37.7 -> 38.5 | Improved |
| add static contexts only: effect on temporal consistency (ablation) | SegFormer MiT-B1 · VSPW val (ablation) | mVC16: 79.9 -> 79.4 | No clear effect |
| add global temporal contexts to CFFM (CFFM++) (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 38.5 -> 39.9 | Improved |
| add global temporal contexts alone to baseline (ablation) | SegFormer MiT-B1 · VSPW val (ablation) | mIoU: 36.5 -> 38.4 | Improved |
| add CFFA only with MLP (no CFM attention) (ablation) | SegFormer MiT-B1 · VSPW val (ablation) | mIoU: 36.5 -> 37.6 | Improved |
| add CFM attention on top of CFFA (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 37.6 -> 38.5 | Improved |
| increase receptive fields from {7,5,3,7} to {49,25,15,7} (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 37.2 -> 39.2 | Improved |
| enlarge receptive fields further to {49,35,21,7} (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 39.2 -> 38.8 | No clear effect |
| use more reference frames {9,6,3} instead of {3,2,1} (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 37.7 -> 38.5 | Improved |
| increase attention layers N from 1 to 4 (ablation) | CFFM MiT-B1 · VSPW val (ablation) | mIoU: 37.8 -> 38.8 | Improved |
| increase prototypes from 10 to 100 (ablation) | CFFM++ MiT-B1 · VSPW val (ablation) | mIoU: 39.5 -> 39.9 | Improved |
Closest papers by meaning
- Coarse-To-Fine Feature Mining for Video Semantic Segmentation (CVPR 2022 · similarity 0.95)
- Mining Relations among Cross-Frame Affinities for Video Semantic Segmentation (ECCV 2022 · similarity 0.88)
- Temporally Grounding Language Queries in Videos by Contextual Boundary-Aware Prediction (AAAI 2020 · similarity 0.88)
- Exploring Temporal Feature Correlation for Efficient and Stable Video Semantic Segmentation (AAAI 2024 · similarity 0.87)
- Exploiting Temporal State Space Sharing for Video Semantic Segmentation (CVPR 2025 · similarity 0.86)
- Re-thinking Temporal Search for Long-Form Video Understanding (CVPR 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.
