Research map / Video / Action recognition
DETACH : Decomposed Spatio-Temporal Alignment for Exocentric Video and Ambient Sensors with Staged Learning
Junho Yoon, Jaemo Jeong, Hyunju Kim, Dongman Lee
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 spatial hard-negative weight W_spatial (ablation) | DETACH · Opportunity++ | weighted F1: 0.56 -> 0.73 | Improved |
| add spatial hard-negative weight W_spatial (ablation) | DETACH · HWU-USP | weighted F1: 0.65 -> 0.71 | Improved |
| add temporal false-negative weight W_temporal (ablation) | DETACH · Opportunity++ | weighted F1: 0.62 -> 0.73 | Improved |
| add temporal false-negative weight W_temporal (ablation) | DETACH · HWU-USP | weighted F1: 0.60 -> 0.71 | Improved |
| replace InfoNCE with spatial-temporal weighted contrastive loss (ablation) | DETACH · HWU-USP | weighted F1: 0.49 -> 0.71 | Improved |
| replace InfoNCE with spatial-temporal weighted contrastive loss (ablation) | DETACH · Opportunity++ | mAP: 0.73 -> 0.83 | Improved |
| add video-guided refinement loss L_refine (ablation) | DETACH · Opportunity++ | weighted F1: 0.60 -> 0.73 | Improved |
| add video-guided refinement loss L_refine (ablation) | DETACH · HWU-USP | weighted F1: 0.69 -> 0.71 | Improved |
| use weighted contrastive loss instead of InfoNCE without refinement (ablation) | DETACH w/o L_refine · Opportunity++ | weighted F1: 0.59 -> 0.60 | No clear effect |
| add momentum encoder for temporal similarity (ablation) | DETACH · Opportunity++ | weighted F1: 0.59 -> 0.73 | Improved |
| add momentum encoder for temporal similarity (ablation) | DETACH · HWU-USP | mAP: 0.61 -> 0.67 | Improved |
| filter to 75th-percentile confident samples instead of all samples (ablation) | DETACH · Opportunity++ | weighted F1: 0.53 -> 0.73 | Improved |
Closest papers by meaning
- EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity (NeurIPS 2024 · similarity 0.87)
- Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives (CVPR 2024 · similarity 0.86)
- Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos (ICLR 2025 · similarity 0.86)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.86)
- Ego4o: Egocentric Human Motion Capture and Understanding from Multi-Modal Input (CVPR 2025 · similarity 0.86)
- Sound Bridge: Associating Egocentric and Exocentric Videos via Audio Cues (CVPR 2025 · 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.
