Research map / Video / Action recognition
Synchronization is All You Need: Exocentric-to-Egocentric Transfer for Temporal Action Segmentation with Unlabeled Synchronized Video Pairs
Camillo Quattrocchi, Antonino Furnari, Daniele Di Mauro, Mario Valerio Giuffrida, Giovanni Maria Farinella
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 |
|---|---|---|---|
| Add TAS model distillation with synchronized exo-ego pairs (DINOv2) (ablation) | Exo-trained C2F-TCN · Assembly101 v3->e4 | Edit: 12.60 -> 28.59 | Improved |
| Add feature distillation only (ablation) | TSM features · Assembly101 v3->e4 | Edit: 10.53 -> 12.38 | Improved |
| Add TAS distillation to TSM (ablation) | TSM features · Assembly101 v3->e4 | Edit: 10.53 -> 20.17 | Improved |
| Add feature distillation on top of TAS distillation (TSM) (ablation) | TSM features · Assembly101 v3->e4 | Edit: 20.17 -> 25.39 | Improved |
| Add feature distillation on top of TAS distillation (DINOv2) (ablation) | DINOv2 features · Assembly101 v3->e4 | Edit: 28.59 -> 28.13 | No clear effect |
| Add feature distillation to TAS distillation on weaker exo view v4 (ablation) | DINOv2 features · Assembly101 v4->e4 | Edit: 24.23 -> 24.46 | No clear effect |
| Use synchronized pairs instead of unpaired adaptation losses (DCORAL, TCC) | TAS distillation · Assembly101 v3->e4 | Edit: 17.84 -> 28.59 | Improved |
| Use random frame pairing (Rnd. MSE) instead of synchronized pairs (ablation) | TAS distillation · Assembly101 v3->e4 | Edit: 12.60 -> 13.27 | No clear effect |
| Reduce adaptation pairs from 100% to 50% (ablation) | TAS distillation · Assembly101 v3->e4 | Edit: 28.59 -> 25.06 | Hurt |
| Randomly drop 3.0% of frames to desynchronize pairs (ablation) | TAS distillation · Assembly101 v3->e4 | Edit: 28.59 -> 22.19 | Hurt |
| Add TAS distillation for EXO->EGO on EgoExo4D | C2F-TCN · EgoExo4D | Edit: 20.56 -> 23.88 | Improved |
| Add TAS distillation for EGO->EXO on EgoExo4D | C2F-TCN · EgoExo4D | MoF: 26.82 -> 25.67 | Mixed |
Closest papers by meaning
- Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos (ICLR 2025 · similarity 0.87)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.87)
- Robust Egocentric Referring Video Object Segmentation via Dual-Modal Causal Intervention (NeurIPS 2025 · similarity 0.87)
- Exocentric-to-Egocentric Video Generation (NeurIPS 2024 · similarity 0.87)
- Progress-Aware Online Action Segmentation for Egocentric Procedural Task Videos (CVPR 2024 · similarity 0.87)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.86)
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.
