Research map / Video / Video segmentation and tracking
Breaking the "Object" in Video Object Segmentation
Read the paperCheck 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 |
|---|---|---|---|
| train AOT on 15 frames instead of 5 (ablation) | AOT · VOST val | J_tr: 36.4 -> 37.4 | Helped |
| replace AOT short-term memory with a recurrent transformer (R-STM) (ablation) | AOT · VOST val | J_tr: 36.4 -> 38.5 | Helped |
| evaluate at 10 fps after adding R-STM (ablation) | AOT + R-STM · VOST val | J_tr: 38.5 -> 40.7 | Helped |
| add multi-scale inference after the 10 fps setting (ablation) | AOT + R-STM + 10 fps · VOST val | J_tr: 40.7 -> 40.1 | Hurt |
| add multi-scale inference after the 10 fps setting (ablation) | AOT + R-STM + 10 fps · VOST val | J: 51.9 -> 52.3 | Helped |
| use in-domain VISOR pre-training before VOST fine-tuning (ablation) | AOT+ · VOST | J_tr | No clear effect |
| add YTVOS video pre-training to static-image pre-training (ablation) | AOT+ · VOST | J_tr | No clear effect |
| increase the VOST training set size (ablation) | AOT+ · VOST val | J_tr | Helped |
| use appearance-based matching on transformed objects | VOS methods · VOST | J_tr | Hurt |
| use low-fps interpolated labels instead of dense annotation for transformations | VOS training data · object transformations | mask annotation quality | Hurt |
Closest papers by meaning
- M^3-VOS: Multi-Phase, Multi-Transition, and Multi-Scenery Video Object Segmentation (CVPR 2025 · similarity 0.92)
- HODOR: High-Level Object Descriptors for Object Re-Segmentation in Video Learned From Static Images (CVPR 2022 · similarity 0.91)
- Test-time Training for Matching-based Video Object Segmentation (NeurIPS 2023 · similarity 0.90)
- Point-VOS: Pointing Up Video Object Segmentation (CVPR 2024 · similarity 0.90)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.89)
- Putting the Object Back into Video Object Segmentation (CVPR 2024 · 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.