Research map / Video / Video segmentation and tracking
DVIS: Decoupled Video Instance Segmentation Framework
Tao Zhang, Xingye Tian, Yu Wu, Shunping Ji, Xuebo Wang, Yuan Zhang, Pengfei Wan
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 |
|---|---|---|---|
| replace standard cross-attention with referring cross-attention (ID added) in tracker (ablation) | DVIS referring tracker · OVIS val (R50, 360p) | AP: 2.9 -> 30.5 | Improved |
| add referring tracker to MinVIS baseline (ablation) | MinVIS · OVIS val (Swin-L, 720p) | AP: 41.6 -> 47.1 | Improved |
| add temporal refiner after tracker (ablation) | DVIS · OVIS val (Swin-L, 720p) | AP: 47.1 -> 49.9 | Improved |
| add referring tracker on lightly occluded objects (ablation) | MinVIS · OVIS val APl | AP (light occlusion): 64.8 -> 64.7 | No clear effect |
| use Hungarian-matched segmenter queries as tracker initial value instead of zero (ablation) | referring tracker · OVIS val (R50, 360p) | AP: 28.9 -> 30.5 | Improved |
| use previous-frame tracker query as initial value instead of matched Q_seg (ablation) | referring tracker · OVIS val (R50, 360p) | AP: 30.5 -> 28.3 | Hurt |
| add Hungarian matching of segmenter queries (ablation) | referring tracker · OVIS val (R50, 360p) | AP: 29.8 -> 30.5 | Improved |
| add long-term temporal attention in refiner (ablation) | temporal refiner · OVIS val (R50) | AP: 31.0 -> 32.2 | Improved |
| add short-term temporal convolution in refiner (ablation) | temporal refiner · OVIS val (R50) | AP: 31.8 -> 32.2 | Improved |
| add cross-attention in refiner (ablation) | temporal refiner · OVIS val (R50) | AP: 30.6 -> 32.2 | Improved |
| increase semi-online inference clip length from 1 to video length (ablation) | DVIS refiner · OVIS val (R50) | AP: 30.0 -> 33.8 | Improved |
| increase refiner training clip length beyond 21 (ablation) | temporal refiner · OVIS val (R50) | AP: 33.8 -> 33.6 | No clear effect |
Later papers that compared with this method
10 later accepted papers report DVIS++ as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- VidEoMT: Your ViT is Secretly Also a Video Segmentation Model (CVPR 2026 · better on 4 of 5 shared results)
- ELVIS: Enhance Low-Light for Video Instance Segmentation in the Dark (CVPR 2026 · better on 4 of 4 shared results)
- CAVIS: Context-Aware Video Instance Segmentation (ICCV 2025 · better on 12 of 12 shared results)
- Temporal-aware Query Routing for Real-time Video Instance Segmentation (ICCV 2025 · better on 6 of 6 shared results)
- LOMM: Latest Object Memory Management for Temporally Consistent Video Instance Segmentation (ICCV 2025 · better on 4 of 4 shared results)
- DVIS++: Improved Decoupled Framework for Universal Video Segmentation (TPAMI 2025 · better on 3 of 3 shared results)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · better on 0 of 1 shared result)
- Improving Video Segmentation via Dynamic Anchor Queries (ECCV 2024 · better on 10 of 12 shared results)
2 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MinVIS — MinVIS: A Minimal Video Instance Segmentation Framework without Video-based Training (NeurIPS 2022) · better on 2 of 2 shared results
- Tube-Link — Tube-Link: A Flexible Cross Tube Framework for Universal Video Segmentation (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DVIS++: Improved Decoupled Framework for Universal Video Segmentation (TPAMI 2025 · similarity 0.94)
- MinVIS: A Minimal Video Instance Segmentation Framework without Video-based Training (NeurIPS 2022 · similarity 0.91)
- Efficient Video Instance Segmentation via Tracklet Query and Proposal (CVPR 2022 · similarity 0.90)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.90)
- Decoupled Motion Expression Video Segmentation (CVPR 2025 · similarity 0.89)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.88)
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.
