Research map / Video / Video segmentation and tracking
Locality-Aware Inter- and Intra-Video Reconstruction for Self-Supervised Correspondence Learning
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 inter-and intra-video reconstruction with 1,440 negative samples (ablation) | LIIR correspondence learning · DAVIS17 val | J&F (Mean): 69.2 -> 72.1 | Helped |
| Add inter-and intra-video reconstruction with 1,440 negative samples (ablation) | LIIR correspondence learning · VIP val | mIoU: 38.4 -> 41.2 | Helped |
| Add 1DAPE position encoding (ablation) | LIIR · DAVIS17 val | J&F (Mean): +1.2 | Helped |
| Use 2DSPE instead of 1DAPE (ablation) | LIIR · DAVIS17 val | J&F (Mean): 70.9 -> 70.6 | Hurt |
| Use 2DAPE instead of 1DAPE (ablation) | LIIR · DAVIS17 val | J&F (Mean): 72.1 -> 71.9 | Hurt |
| Shift position encoding for inter-video matching rather than leave it unmodified (ablation) | LIIR · DAVIS17 val | J&F (Mean): 71.3 -> 72.1 | Helped |
| Shift position encoding rather than shuffle it (ablation) | LIIR · DAVIS17 val | J&F (Mean): 71.8 -> 72.1 | Helped |
| Add spatial compactness during training and inference (ablation) | LIIR · DAVIS17 val | J&F (Mean): 69.8 -> 72.1 | Helped |
| Add spatial compactness during training only (ablation) | LIIR · DAVIS17 val | J&F (Mean): +1.7 | Helped |
| Add spatial compactness during inference only (ablation) | LIIR · DAVIS17 val | J&F (Mean): +1.0 | Helped |
| Use one Gaussian component for compactness during training rather than two (ablation) | LIIR spatial compactness prior · DAVIS17 val | J&F (Mean): 69.8 -> 69.4 | Hurt |
| Use two Gaussian components rather than three for training compactness (ablation) | LIIR spatial compactness prior · DAVIS17 val | J&F (Mean): 71.3 -> 71.5 | Helped |
Closest papers by meaning
- Joint-task Self-supervised Learning for Temporal Correspondence (NeurIPS 2019 · similarity 0.92)
- When the Future Becomes the Past: Taming Temporal Correspondence for Self-supervised Video Representation Learning (CVPR 2025 · similarity 0.87)
- Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency (CVPR 2025 · similarity 0.87)
- FRESCO: Spatial-Temporal Correspondence for Zero-Shot Video Translation (CVPR 2024 · similarity 0.85)
- Emergent Temporal Correspondences from Video Diffusion Transformers (NeurIPS 2025 · similarity 0.85)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.84)
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.