Research map / Video / Video segmentation and tracking
F2Net: Learning to Focus on the Foreground for Unsupervised 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 |
|---|---|---|---|
| Add center-point Gaussian spatial guidance to appearance diffusion (ablation) | Appearance diffusion on DAVIS2016 · DAVIS2016 | J Mean: 79.6 -> 82.5 | Helped |
| Add center-point Gaussian spatial guidance to appearance diffusion (ablation) | Appearance diffusion on DAVIS2016 · DAVIS2016 | F Mean: 79.0 -> 82.9 | Helped |
| Add ordinary appearance diffusion to the DeepLabV3 baseline (ablation) | DeepLabV3 baseline on DAVIS2016 · DAVIS2016 | J Mean: 77.6 -> 79.6 | Helped |
| Add ordinary appearance diffusion to the DeepLabV3 baseline (ablation) | DeepLabV3 baseline on DAVIS2016 · DAVIS2016 | F Mean: 76.8 -> 79.0 | Helped |
| Add spatial-only attention to CGAD (ablation) | CGAD on DAVIS2016 · DAVIS2016 | J Mean: +0.2 | Helped |
| Add spatial-only attention to CGAD (ablation) | CGAD on DAVIS2016 · DAVIS2016 | F Mean: +0.5 | Helped |
| Add channel-only attention to CGAD (ablation) | CGAD on DAVIS2016 · DAVIS2016 | J Mean: +0.3 | Helped |
| Add channel-only attention to CGAD (ablation) | CGAD on DAVIS2016 · DAVIS2016 | F Mean: +0.7 | Helped |
| Use channel attention followed by spatial attention rather than spatial followed by channel attention (ablation) | CGAD on DAVIS2016 · DAVIS2016 | J Mean: 83.0 -> 83.1 | Helped |
| Use channel attention followed by spatial attention rather than spatial followed by channel attention (ablation) | CGAD on DAVIS2016 · DAVIS2016 | F Mean: 84.2 -> 84.4 | Helped |
| Use motion history to select the center candidate instead of choosing the heatmap maximum (ablation) | F2Net on DAVIS2016 · DAVIS2016 | J Mean: 82.4 -> 83.1 | Helped |
| Use motion history to select the center candidate instead of choosing the heatmap maximum (ablation) | F2Net on DAVIS2016 · DAVIS2016 | F Mean: 84.2 -> 84.4 | Helped |
Closest papers by meaning
- See More, Know More: Unsupervised Video Object Segmentation With Co-Attention Siamese Networks (CVPR 2019 · similarity 0.88)
- FOCUS: Towards Universal Foreground Segmentation (AAAI 2025 · similarity 0.87)
- Target-Aware Object Discovery and Association for Unsupervised Video Multi-Object Segmentation (CVPR 2021 · similarity 0.87)
- Video K-Net: A Simple, Strong, and Unified Baseline for Video Segmentation (CVPR 2022 · similarity 0.87)
- Guided Slot Attention for Unsupervised Video Object Segmentation (CVPR 2024 · similarity 0.86)
- Dual Prototype Attention for Unsupervised Video Object Segmentation (CVPR 2024 · similarity 0.85)
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.