Research map / Video / Video segmentation and tracking
In Defense of Online Models for Video Instance Segmentation
Junfeng Wu, Qihao Liu, Yi Jiang, Song Bai, Alan Yuille, Xiang Bai
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 contrastive training instead of ID head, with multi-cues matching (ablation) | IDOL · YouTube-VIS 2019 val | AP: 30.3 -> 31.8 | Improved |
| add contrastive training instead of ID head, with embedding matching (ablation) | IDOL · YouTube-VIS 2019 val | AP: 33.6 -> 42.5 | Improved |
| add contrastive training instead of ID head (ablation) | IDOL · OVIS val | AP: 11.0 -> 18.4 | Improved |
| add optimal transport sample selection (ablation) | IDOL · YouTube-VIS 2019 val | AP: 42.5 -> 44.8 | Improved |
| add optimal transport sample selection (ablation) | IDOL · OVIS val | AP: 26.7 -> 28.3 | Improved |
| use embedding association instead of multi-cues (ablation) | IDOL · YouTube-VIS 2019 val | AP: 31.8 -> 42.5 | Improved |
| use embedding association instead of multi-cues (ablation) | IDOL · OVIS val | AP: 18.4 -> 26.7 | Improved |
| add temporally weighted softmax (ablation) | IDOL · YouTube-VIS 2019 val | AP: 44.8 -> 46.4 | Improved |
| add temporally weighted softmax (ablation) | IDOL · OVIS val | AP: 28.3 -> 30.2 | Improved |
| pretrain contrastive embedding on COCO pseudo key-reference pairs | IDOL ResNet-50 · YouTube-VIS 2019 val | AP: 46.4 -> 49.5 | Improved |
| pretrain contrastive embedding on COCO pseudo key-reference pairs | IDOL Swin-L · YouTube-VIS 2019 val | AP: 61.5 -> 64.3 | Improved |
| increase clip length from 1 to 30 (frame oracle) (ablation) | SeqFormer · OVIS val | AP: 32.0 -> 8.9 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SeqFormer — SeqFormer: Sequential Transformer for Video Instance Segmentation (ECCV 2022) · better on 3 of 3 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
- InstanceFormer: An Online Video Instance Segmentation Framework (AAAI 2023 · similarity 0.93)
- MiNQVIS: Mitigating Noisy Queries for Robust Online Video Instance Segmentation (ECCV 2026 · similarity 0.92)
- CTVIS: Consistent Training for Online Video Instance Segmentation (ICCV 2023 · similarity 0.92)
- InsPro: Propagating Instance Query and Proposal for Online Video Instance Segmentation (NeurIPS 2022 · similarity 0.91)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.90)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.89)
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.
