Research map / Video / Video segmentation and tracking
Temporally Distributed Networks for Fast Video Semantic Segmentation
Check 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 |
|---|---|---|---|
| Distribute computation across two independent feature paths and aggregate them over time | PSPNet101-based video segmentation · Cityscapes | mIoU (%): 79.2 -> 79.4 | Helped |
| Distribute computation across two independent feature paths and aggregate them over time | PSPNet101-based video segmentation · CamVid | mIoU (%): 76.2 -> 76.0 | No clear effect |
| Distribute computation across two independent feature paths and aggregate them over time | PSPNet101-based video segmentation · NYUD-v2 | mIoU (%): 43.2 -> 43.5 | Helped |
| Add grouped knowledge distillation to whole-model distillation (ablation) | TD⁴-PSP18 · Cityscapes | mIoU (%): 76.5 -> 76.8 | Helped |
| Add whole-model knowledge distillation to supervised training (ablation) | TD⁴-PSP18 · Cityscapes | mIoU (%): 76.4 -> 76.5 | Helped |
| Increase attention downsampling stride from n=1 to n=4 (ablation) | TD²-PSP50 · Cityscapes | mIoU (%): 80.0 -> 79.9 | No clear effect |
| Increase attention downsampling stride from n=4 to n=32 (ablation) | TD²-PSP50 · Cityscapes | mIoU (%): 79.9 -> 79.1 | Hurt |
| Use independent subnetworks instead of one shared subnetwork across paths (ablation) | TDNet · Cityscapes | mIoU (%): 78.5 -> 79.9 | Helped |
| Use four shallow PSPNet18 paths rather than two PSPNet50 paths (ablation) | TDNet · Cityscapes | mIoU (%): 79.9 -> 76.8 | Hurt |
| Use four shallow PSPNet18 paths rather than two PSPNet50 paths (ablation) | TDNet · CamVid | mIoU (%): 76.0 -> 72.6 | Hurt |
| Use four shallow PSPNet18 paths rather than two PSPNet50 paths (ablation) | TDNet · NYUD-v2 | mIoU (%): 43.5 -> 37.4 | Hurt |
| Add attention-based motion compensation rather than simply adding feature groups (ablation) | TD⁴-PSP18 · Cityscapes | mIoU (%) | Helped |
Closest papers by meaning
- Low-Latency Video Semantic Segmentation (CVPR 2018 · similarity 0.90)
- Dynamic Video Segmentation Network (CVPR 2018 · similarity 0.90)
- Exploiting Temporal State Space Sharing for Video Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Pay Attention to Target: Relation-Aware Temporal Consistency for Domain Adaptive Video Semantic Segmentation (AAAI 2024 · similarity 0.87)
- SCTNet: Single-Branch CNN with Transformer Semantic Information for Real-Time Segmentation (AAAI 2024 · similarity 0.86)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (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.