Research map / Video / Video segmentation and tracking
VISAGE: Video Instance Segmentation with Appearance-Guided Enhancement
Hanjung Kim, Jaehyun Kang, Miran Heo, Sukjun Hwang, Seoung Wug Oh, Seon Joo Kim
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 appearance guidance to matching (with memory bank) (ablation) | VISAGE · YouTube-VIS 2019 val | AP: 53.4 -> 55.1 | Improved |
| add memory bank (with appearance guidance) (ablation) | VISAGE · YouTube-VIS 2019 val | AP: 50.2 -> 55.1 | Improved |
| build appearance queries from backbone features rather than transformer encoder features (ablation) | VISAGE · YouTube-VIS 2019 val | AP: 51.4 -> 55.1 | Improved |
| use simplified tracker instead of prior tracker with NMS heuristics (ablation) | VISAGE · YouTube-VIS 2019 val | AP: 54.4 -> 55.1 | Improved |
| add NMS to the prior tracker (ablation) | prior tracker · YouTube-VIS 2019 val | AP: 52.7 -> 54.4 | Improved |
| use appearance weight 0.75 instead of 0 (location only) (ablation) | VISAGE · OVIS val | AP: 32.2 -> 36.2 | Improved |
| rely only on appearance (alpha=1) instead of mix (ablation) | VISAGE · YouTube-VIS 2019 val | AP: 55.1 -> 24.9 | Hurt |
| add appearance information to matching (ablation) | VISAGE · Pseudo dataset (swap) | AP: 51.8 -> 63.0 | Improved |
| add appearance information to matching (ablation) | VISAGE · Pseudo dataset (track) | AP: 65.2 -> 65.8 | Improved |
| VISAGE vs CTVIS on YouTube-VIS 2019 | VISAGE · YouTube-VIS 2019 val | AP: 55.1 vs 55.1 | No clear effect |
Later papers that compared with this method
3 later accepted papers report ViSAGe as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Towards Streaming Synchronized Spatial Audio Generation via Autoregressive Diffusion Transformer (ICML 2026 · better on 2 of 2 shared results)
- CAVIS: Context-Aware Video Instance Segmentation (ICCV 2025 · better on 3 of 3 shared results)
- OmniAudio: Generating Spatial Audio from 360-Degree Video (ICML 2025 · better on 1 of 1 shared result)
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.
- CTVIS — CTVIS: Consistent Training for Online Video Instance Segmentation (ICCV 2023) · better on 4 of 5 shared results
- TCOVIS — TCOVIS: Temporally Consistent Online Video Instance Segmentation (ICCV 2023) · better on 3 of 4 shared results
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
- InsPro: Propagating Instance Query and Proposal for Online Video Instance Segmentation (NeurIPS 2022 · similarity 0.91)
- Efficient Video Instance Segmentation via Tracklet Query and Proposal (CVPR 2022 · similarity 0.90)
- In Defense of Online Models for Video Instance Segmentation (ECCV 2022 · similarity 0.89)
- DVIS: Decoupled Video Instance Segmentation Framework (ICCV 2023 · similarity 0.89)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.88)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.87)
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.
