Research map / Detection and segmentation / Open-set and few-shot detection
Instance Embedding Transfer to 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 |
|---|---|---|---|
| Transfer image-trained instance embeddings and combine them with objectness, optical flow, and seed adaptation | Unsupervised video object segmentation · DAVIS 2016 val | J Mean: 78.5 vs 76.2 | Helped |
| Transfer image-trained instance embeddings and combine them with objectness, optical flow, and seed adaptation | Unsupervised video object segmentation · FBMS test | F-score: 82.8 vs 77.8 | Helped |
| Transfer image-trained instance embeddings and combine them with objectness, optical flow, and seed adaptation | Unsupervised video object segmentation · FBMS test | J Mean: 71.9 vs 59.8 | Helped |
| Use the proposed unsupervised method | Video object segmentation · SegTrack-v2 | J: 59.3% vs 67.2% | Hurt |
| Use instance embeddings instead of DeepLab-v2 fc7 semantic features (ablation) | Proposed segmentation method · DAVIS train | J: 65.2% with fc7; more than 10% less than instance embeddings | Helped |
| Increase online adaptation frequency from none to every frame (ablation) | Proposed segmentation method · DAVIS train | J Mean: 70.5 -> 77.5 | Helped |
| Adapt seed sets every 5 frames instead of every frame (ablation) | Proposed segmentation method · DAVIS train | J Mean: 77.5 -> 76.4 | Hurt |
| Adapt seed sets every 10 frames instead of every frame (ablation) | Proposed segmentation method · DAVIS train | J Mean: 77.5 -> 75.9 | Hurt |
| Adapt seed sets every 20 frames instead of every frame (ablation) | Proposed segmentation method · DAVIS train | J Mean: 77.5 -> 75.0 | Hurt |
| Adapt seed sets every 40 frames instead of every frame (ablation) | Proposed segmentation method · DAVIS train | J Mean: 77.5 -> 73.6 | Hurt |
| Rank foreground seed tracks using both motion saliency and objectness rather than motion alone (ablation) | Proposed segmentation method · DAVIS train | J Mean: 74.3 -> 77.5 | Helped |
| Rank foreground seed tracks using both motion saliency and objectness rather than objectness alone (ablation) | Proposed segmentation method · DAVIS train | J Mean: 75.7 -> 77.5 | Helped |
Closest papers by meaning
- Video Object Segmentation Using Global and Instance Embedding Learning (CVPR 2021 · similarity 0.91)
- Unified Mask Embedding and Correspondence Learning for Self-Supervised Video Segmentation (CVPR 2023 · similarity 0.89)
- Dense Unsupervised Learning for Video Segmentation (NeurIPS 2021 · similarity 0.88)
- Sparse Object-Level Supervision for Instance Segmentation With Pixel Embeddings (CVPR 2022 · similarity 0.87)
- Weakly Supervised Instance Segmentation for Videos With Temporal Mask Consistency (CVPR 2021 · similarity 0.87)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.87)
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.