Research map / Video / Video-language understanding
Egocentric Video-Language Pretraining
Kevin Qinghong Lin, Jinpeng Wang, Mattia Soldan, Michael Wray, Rui Yan, Zhongcong Xu, Difei Gao, Rong-Cheng Tu, Wenzhe Zhao, Weijie Kong, Chengfei Cai, WANG HongFa and 4 more
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 |
|---|---|---|---|
| pretrain on EgoClip instead of CC3M+WebVid-2M | Frozen · EPIC-KITCHENS-100 MIR zero-shot | mAP avg: 8.0 -> 15.5 | Improved |
| pretrain on EgoClip instead of HowTo100M | Frozen · Ego4D NLQ val IoU=0.3 | R@1: 3.95 -> 10.53 | Improved |
| replace InfoNCE with EgoNCE (ablation) | Frozen on EgoClip · EPIC-KITCHENS-100 MIR 4 frames | nDCG avg: 54.3 -> 54.8 | Improved |
| replace InfoNCE with EgoNCE (ablation) | Frozen on EgoClip · Charades-Ego fine-tuned | mAP: 31.2 -> 32.1 | Improved |
| replace InfoNCE with EgoNCE (ablation) | Frozen on EgoClip · Ego4D OSCC val | accuracy: 73.4 -> 73.9 | Improved |
| replace InfoNCE with EgoNCE (ablation) | Frozen on EgoClip · EgoMCQ intra-video | accuracy: 51.5 -> 57.2 | Improved |
| action-aware positives using nouns only (ablation) | InfoNCE · EgoMCQ intra-video | accuracy: 51.5 -> 42.3 | Hurt |
| action-aware positives using noun and verb together (ablation) | InfoNCE · EgoMCQ intra-video | accuracy: 51.5 -> 53.6 | Improved |
| use random negatives (ablation) | InfoNCE · EgoMCQ intra-video | accuracy: 51.5 -> 49.9 | Hurt |
| use hard negatives within 1 minute (ablation) | InfoNCE · EgoMCQ intra-video | accuracy: 51.5 -> 54.5 | Improved |
| center clip at timestamp rather than start at it (ablation) | EgoClip creation · EgoMCQ intra-video | accuracy: 39.72 -> 41.68 | Improved |
| use per-video variable clip length beta_i instead of fixed length (ablation) | EgoClip creation · EgoMCQ intra-video | accuracy: 41.68 -> 44.82 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Ego-Exo — Ego-Exo: Transferring Visual Representations From Third-Person to First-Person Videos (CVPR 2021) · better on 1 of 1 shared result
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
- Do Egocentric Video-Language Models Truly Understand Hand-Object Interactions? (ICLR 2025 · similarity 0.91)
- Retrieval-Augmented Egocentric Video Captioning (CVPR 2024 · similarity 0.91)
- MMEgo: Towards Building Egocentric Multimodal LLMs for Video QA (ICLR 2025 · similarity 0.88)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.88)
- EgoVid-5M: A Large-Scale Video-Action Dataset for Egocentric Videos Generation (NeurIPS 2025 · similarity 0.88)
- X-MIC: Cross-Modal Instance Conditioning for Egocentric Action Generalization (CVPR 2024 · similarity 0.88)
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.
