Research map / Video / Video segmentation and tracking
Learning to Track Instances without Video Annotations
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 an instance contrastive embedding to SOLO (ablation) | SOLO · YouTube-VIS validation | AP: 23.9 -> 28.4 | Helped |
| Add maximum entropy regularization to IC embedding (ablation) | SOLO-Track · YouTube-VIS validation | AP: 28.4 -> 29.7 | Helped |
| Add self-supervised video correspondence (ablation) | SOLO-Track with IC and ME · YouTube-VIS validation | AP: 29.7 -> 32.9 | Helped |
| Increase cycle-loss sequence length from 2 to 3 frames (ablation) | Video correspondence model · YouTube-VIS validation | AP: 32.9 -> 31.8 | Hurt |
| Increase cycle-loss sequence length from 2 to 4 frames (ablation) | Video correspondence model · YouTube-VIS validation | AP: 32.9 -> 30.9 | Hurt |
| Add test-time adaptation using self-supervised video correspondence | Ours+ · YouTube-VIS validation after post-processing | AP: 37.4 -> 38.3 | Helped |
| Use the proposed tracking embedding rather than OKS-only association | PoseTrack baseline · PoseTrack2018 validation | MOTA Total: 62.0 -> 62.8 | Helped |
| Add video correspondence to the pose tracking model | Ours · PoseTrack2018 validation | MOTA Total: 62.8 -> 64.2 | Helped |
| Fine-tune the pose model with MPII data (ablation) | Ours+ · PoseTrack2018 validation | AP: 69.3 -> 71.4 | Helped |
| Replace joint-level IC loss with Associate Embedding loss (ablation) | Pose tracking model · PoseTrack2018 validation | MOTA: 64.7 -> 63.5 | Hurt |
Closest papers by meaning
- Weakly Supervised Instance Segmentation for Videos With Temporal Mask Consistency (CVPR 2021 · similarity 0.91)
- TrackIME: Enhanced Video Point Tracking via Instance Motion Estimation (NeurIPS 2024 · similarity 0.89)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.89)
- Matching Anything by Segmenting Anything (CVPR 2024 · similarity 0.88)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.88)
- Is ImageNet worth 1 video? Learning strong image encoders from 1 long unlabelled video (ICLR 2024 · 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.