Research map / Video / Video-language understanding
EgoDistill: Egocentric Head Motion Distillation for Efficient Video Understanding
Shuhan Tan, Tushar Nagarajan, Kristen Grauman
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 |
|---|---|---|---|
| remove L_KD (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 39.33 | Hurt |
| remove L_1 feature-matching loss (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 42.20 | Hurt |
| remove L_GT (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 43.17 | Hurt |
| remove L_KD (ablation) | EgoDistill · Ego4D | top-1 accuracy: 37.95 -> 35.51 | Hurt |
| remove self-supervised IMU pretraining (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 41.21 | Hurt |
| remove L_1 pretraining (keep L_NCE) (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 42.30 | Hurt |
| remove L_NCE pretraining (keep L_1) (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 43.51 | Hurt |
| add IMU and distillation to image model trained with L_GT only (ablation) | VisOnly · EPIC-Kitchens | top-1 accuracy: 35.04 -> 44.95 | Improved |
| add IMU branch to VisOnly-Distill | VisOnly-Distill (MotionFormer teacher) · EPIC-Kitchens | top-1 accuracy: 37.20 -> 44.95 | Improved |
| add IMU branch to VisOnly-Distill | VisOnly-Distill (SlowFast teacher) · Ego4D | top-1 accuracy: 33.04 -> 33.29 | Improved |
| use MViT teacher instead of MotionFormer teacher | EgoDistill · EPIC-Kitchens | top-1 accuracy: 44.95 -> 36.90 | Hurt |
| uniform vs last-frame sampling (N=4) (ablation) | EgoDistill · EPIC-Kitchens | top-1 accuracy: 41.72 -> 52.43 | Improved |
Later papers that compared with this method
1 later accepted paper reports EgoDistill as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- SkillSight: Efficient First-Person Skill Assessment with Gaze (CVPR 2026 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Egocentric Cues from Exocentric Video using Privileged Egocentric Supervision (ECCV 2026 · similarity 0.89)
- Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos (ICLR 2025 · similarity 0.88)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.87)
- Retrieval-Augmented Egocentric Video Captioning (CVPR 2024 · similarity 0.87)
- MMEgo: Towards Building Egocentric Multimodal LLMs for Video QA (ICLR 2025 · similarity 0.87)
- EgoVid-5M: A Large-Scale Video-Action Dataset for Egocentric Videos Generation (NeurIPS 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.
