Research map / Video / Video segmentation and tracking
Unified Mask Embedding and Correspondence Learning for Self-Supervised Video 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 mask-embedded segmentation learning to short- and long-term correspondence learning (ablation) | Unified framework · DAVIS17 val | J&F_m: 68.8 -> 74.5 | Helped |
| Add correspondence learning to segmentation-only training (ablation) | Unified framework · DAVIS17 val | J&F_m: 62.3 -> 74.5 | Helped |
| Combine short- and long-term correspondence losses (ablation) | Correspondence learning · DAVIS17 val | J&F_m: 57.4 -> 68.8 | Helped |
| Use recurrent mask refinement (ablation) | Unified framework · DAVIS17 val | J&F_m: 69.7 -> 74.5 | Helped |
| Update pseudo masks every 10 epochs rather than leaving them fixed (ablation) | Unified framework · DAVIS17 val | J&F_m: 71.1 -> 74.5 | Helped |
| Update pseudo masks every epoch rather than every 10 epochs (ablation) | Unified framework · DAVIS17 val | J&F_m: 74.5 -> 69.7 | Hurt |
| Use 5 cluster centers rather than 2, 3, 8, or 10 (ablation) | Space-time clustering · DAVIS17 val | J&F_m: 67.5 / 71.6 / 72.5 / 70.1 -> 74.5 | Helped |
| Use MAST reconstruction plus segmentation learning (ablation) | MAST · DAVIS17 val | J&F_m: +3.5 | Helped |
| Use CRW cycle-consistency plus segmentation learning (ablation) | CRW · DAVIS17 val | J&F_m: +4.2 | Helped |
| Use the proposed contrastive correspondence loss plus segmentation learning (ablation) | Unified framework · DAVIS17 val | J&F_m: +5.7 | Helped |
| Use the full method with ResNet-50 instead of ResNet-18 | Unified framework · YouTube-VOS val | J&F_m: 71.6 -> 72.4 | Helped |
Closest papers by meaning
- Boosting Video Object Segmentation via Space-Time Correspondence Learning (CVPR 2023 · similarity 0.92)
- Self-supervised Object-Centric Learning for Videos (NeurIPS 2023 · similarity 0.91)
- Video Object Segmentation Using Global and Instance Embedding Learning (CVPR 2021 · similarity 0.91)
- Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation (CVPR 2020 · similarity 0.90)
- Mask-Free Video Instance Segmentation (CVPR 2023 · similarity 0.89)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · 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.