Research map / Video / Video segmentation and tracking
Generating Masks From Boxes by Mining Spatio-Temporal Consistencies in Videos
Bin Zhao, Goutam Bhat, Martin Danelljan, Luc Van Gool, Radu Timofte
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 |
|---|---|---|---|
| use 9 input frames instead of 1 (ablation) | box-to-mask network · DAVIS 2017 val | Jaccard J: 78.7 -> 81.2 | Improved |
| use 9 input frames instead of 1 (ablation) | box-to-mask network · YT300 | Jaccard J: 84.2 -> 85.6 | Improved |
| increase frames beyond 9 to 11 (ablation) | box-to-mask network · DAVIS 2017 val | Jaccard J: 81.2 -> 81.2 | No clear effect |
| add spatio-temporal aggregation module (MultiFrame) over SingleImage (ablation) | box-to-mask network · DAVIS 2017 val | Jaccard J: 77.2 -> 79.8 | Improved |
| add single-frame representation to the decoder (ablation) | MultiFrame · DAVIS 2017 val | Jaccard J: 79.8 -> 80.1 | Improved |
| add iterative refinement (ablation) | MultiFrame+ · DAVIS 2017 val | Jaccard J: 80.1 -> 81.2 | Improved |
| add iterative refinement (ablation) | MultiFrame+ · YT300 | Jaccard J: 84.8 -> 85.6 | Improved |
| use more steepest-descent iterations at inference (5 to 15) (ablation) | aggregation module · DAVIS 2017 val | Jaccard J: 80.0 -> 81.2 | Improved |
| use more than 15 iterations (15 to 20) (ablation) | aggregation module · DAVIS 2017 val | Jaccard J: 81.2 -> 81.2 | No clear effect |
| increase sample scale from 2 to 3 (ablation) | box-to-mask network · YT300 | Jaccard J: 84.5 -> 85.7 | Improved |
| increase sample scale from 4 to 5 (ablation) | box-to-mask network · DAVIS 2017 val | Jaccard J: 81.2 -> 80.8 | Hurt |
| use inter-frame interval 5 instead of consecutive frames (ablation) | box-to-mask network · DAVIS 2017 val | Jaccard J: 80.5 -> 81.2 | Improved |
Closest papers by meaning
- Track and Segment: An Iterative Unsupervised Approach for Video Object Proposals (CVPR 2016 · similarity 0.90)
- Boosting Video Object Segmentation via Space-Time Correspondence Learning (CVPR 2023 · similarity 0.90)
- Weakly Supervised Instance Segmentation for Videos With Temporal Mask Consistency (CVPR 2021 · similarity 0.89)
- Robust Visual Tracking by Segmentation (ECCV 2022 · similarity 0.89)
- Multi-Granularity Video Object Segmentation (AAAI 2025 · similarity 0.88)
- Point-VOS: Pointing Up 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.
