Research map / Video / Video segmentation and tracking
Less than Few: Self-Shot Video Instance Segmentation
Pengwan Yang, Yuki M. Asano, Pascal Mettes, Cees G. M. Snoek
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 |
|---|---|---|---|
| retrieve support videos with self-shot Video Rank features instead of random videos | VIS transformer · Self-VIS test (1 support video) | mAP@0.5: 44.3 -> 51.4 | Improved |
| use Video Rank loss instead of MIL-NCE for retrieval | self-shot retrieval · Self-VIS test (1 support video) | mAP@0.5: 50.1 -> 51.4 | Improved |
| finetune Inception features with MoCo instead of fixed | self-shot retrieval · Self-VIS test (1 support video) | mAP@0.5: 46.9 -> 49.4 | Improved |
| add 5 extra support videos at inference to 5 self-shot supports | Video Rank self-shot · Self-VIS test (5 support videos) | mAP@0.5: 54.3 -> 55.4 | Improved |
| use self-shot support instead of zero-shot | VIS transformer · Self-VIS | mAP@0.5: 47.4 -> 51.4 | Improved |
| use self-shot support instead of zero-shot | temporal action localization · Thumos14 | video-mAP@0.5: 43.4 -> 45.8 | Improved |
| use unlabelled pool supports instead of segments of query video | self-shot · Self-VIS test (1 support video) | mAP@0.5: 44.9 -> 51.4 | Improved |
| add the query-support fuser (ablation) | VIS transformer decoder · Self-VIS (1 self-shot) | mAP@0.5: 40.7 -> 49.1 | Improved |
| add the transformer encoder (ablation) | VIS transformer decoder · Self-VIS (1 self-shot) | mAP@0.5: 40.7 -> 42.4 | Improved |
| add the encoder on top of fuser and decoder (ablation) | VIS transformer · Self-VIS (1 self-shot) | mAP@0.5: 49.1 -> 51.4 | Improved |
| add the fuser (ablation) | VIS transformer decoder · Self-OVIS (1 self-shot) | mAP@0.5: 13.7 -> 19.6 | Improved |
| add spatio-temporal positional encoding to support and query (ablation) | VIS transformer (oracle one-shot) · Self-VIS | mAP@0.5: 49.4 -> 53.2 | Improved |
Closest papers by meaning
- Boosting Few-Shot Visual Learning With Self-Supervision (ICCV 2019 · similarity 0.90)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.90)
- Incremental Few-Shot Instance Segmentation (CVPR 2021 · similarity 0.90)
- Generalized Few-Shot Video Classification With Video Retrieval and Feature Generation (TPAMI 2021 · similarity 0.89)
- A Survey of Self-Supervised and Few-Shot Object Detection (TPAMI 2022 · similarity 0.89)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · 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.
