Research map / Detection and segmentation / Tracking
Fast Video Object Segmentation With Temporal Aggregation Network and Dynamic Template Matching
Check 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 |
|---|---|---|---|
| Add TAN temporal aggregation (ablation) | Faster R-CNN · DAVIS-2017 validation, 800 resolution | APbbox: 45.3 -> 46.3 | Helped |
| Add TAN temporal aggregation (ablation) | Faster R-CNN · DAVIS-2017 validation, 800 resolution | APmask: 35.4 -> 36.7 | Helped |
| Feed continuous neighboring frames rather than duplicate copies (ablation) | TAN with ResNet-50 · ImageNet VID validation | APbbox50: 76.0 -> 78.2 | Helped |
| Use DTTM instead of IoU-only matching (ablation) | Tracking-by-detection VOS · DAVIS validation | 𝒥&ℱ-Mean: 69.2 -> 71.7 | Helped |
| Replace DTTM with moving-average template updates at mnt=0.3 (ablation) | Tracking-by-detection VOS · DAVIS validation | 𝒥&ℱ-Mean: 69.2 -> 70.8 | Helped |
| Use moving-average template updates instead of DTTM (ablation) | Tracking-by-detection VOS · DAVIS validation | 𝒥&ℱ-Mean: 71.7 -> 70.8 | Hurt |
| Set DTTM appearance threshold to 0.3 instead of 0.5 (ablation) | Tracking-by-detection VOS · DAVIS validation | 𝒥&ℱ-Mean: 71.7 -> 70.5 | Hurt |
| Set DTTM appearance threshold to 0.7 instead of 0.5 (ablation) | Tracking-by-detection VOS · DAVIS validation | 𝒥&ℱ-Mean: 71.7 -> 70.6 | Hurt |
| Add join skip connections to the FCN segmentation head (ablation) | FCN segmentation head · DAVIS validation | 𝒥&ℱ-Mean: 67.9 -> 69.2 | Helped |
| Add cascade skip connections to the FCN segmentation head (ablation) | FCN segmentation head · DAVIS validation | 𝒥&ℱ-Mean: 67.9 -> 68.8 | Helped |
| Use the proposed tracking-by-detection pipeline | Video object segmentation · DAVIS 2019 validation | 𝒥&ℱ-Mean: 75.9 | Helped |
Closest papers by meaning
- State-Aware Tracker for Real-Time Video Object Segmentation (CVPR 2020 · similarity 0.89)
- Boosting Video Object Segmentation via Space-Time Correspondence Learning (CVPR 2023 · similarity 0.89)
- Target-Aware Object Discovery and Association for Unsupervised Video Multi-Object Segmentation (CVPR 2021 · similarity 0.89)
- Test-time Training for Matching-based Video Object Segmentation (NeurIPS 2023 · similarity 0.89)
- Fast Template Matching and Update for Video Object Tracking and Segmentation (CVPR 2020 · similarity 0.88)
- Fast and Accurate Online Video Object Segmentation via Tracking Parts (CVPR 2018 · 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.