Research map / Detection and segmentation / CNN features
Associating Objects with Transformers for Video Object Segmentation
Zongxin Yang, Yunchao Wei, Yi Yang
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 short-term local attention (window 15 vs none) (ablation) | AOT-S · YouTube-VOS 2018 val | J&F: 74.3 -> 80.3 | Improved |
| increase LSTT blocks from 1 to 2 (ablation) | AOT · YouTube-VOS 2018 val | J&F: 77.9 -> 80.3 | Improved |
| increase LSTT blocks from 2 to 3 (ablation) | AOT · YouTube-VOS 2018 val | J&F: 80.3 -> 80.9 | Improved |
| increase identity number M from 10 to 15 (ablation) | AOT-S · YouTube-VOS 2018 val | J&F: 80.3 -> 79.0 | Hurt |
| increase identity number M from 10 to 30 (ablation) | AOT-S · YouTube-VOS 2018 val | J&F: 80.3 -> 77.2 | Hurt |
| use 2 previous frames in local attention instead of 1 (ablation) | AOT-S · YouTube-VOS 2018 val | J&F: 80.3 -> 80.0 | Hurt |
| add relative positional embedding to local attention (ablation) | AOT-S · YouTube-VOS 2018 val | J&F: 79.7 -> 80.3 | Improved |
| add sine positional embedding to self-attention (ablation) | AOT-S · YouTube-VOS 2018 val | J&F: 80.1 -> 80.3 | Improved |
| train with YouTube-VOS data for DAVIS 2017 | AOT-S · DAVIS 2017 val | J&F: 79.2 -> 81.3 | Improved |
| use AOT-S instead of CFBI+ at similar accuracy | AOT-S · YouTube-VOS 2018 val | FPS: 4.0 -> 27.1 | Improved |
| swap R50 encoder for SwinB | AOT-L · YouTube-VOS 2018 val | J&F: 84.1 -> 84.5 | Improved |
Closest papers by meaning
- Scalable Video Object Segmentation With Identification Mechanism (TPAMI 2024 · similarity 0.94)
- Boosting Video Object Segmentation via Space-Time Correspondence Learning (CVPR 2023 · similarity 0.89)
- Decoupling Features in Hierarchical Propagation for Video Object Segmentation (NeurIPS 2022 · similarity 0.89)
- Breaking the "Object" in Video Object Segmentation (CVPR 2023 · similarity 0.88)
- Video Object Segmentation With Dynamic Memory Networks and Adaptive Object Alignment (ICCV 2021 · similarity 0.88)
- Video Object Segmentation Using Global and Instance Embedding Learning (CVPR 2021 · 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.
