Research map / Vision-language and multimodal / Robot action and navigation
EgoAdapt: Adaptive Multisensory Distillation and Policy Learning for Efficient Egocentric Perception
Sanjoy Chowdhury, Subrata Biswas, Sayan Nag, Tushar Nagarajan, Calvin Murdock, Ishwarya Ananthabhotla, Yijun Qian, Vamsi Krishna Ithapu, Dinesh Manocha, Ruohan Gao
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 policy module TeMPLe on top of distillation (ablation) | EgoAdapt · EPIC-Kitchens | action accuracy: 50.88 -> 56.74 | Improved |
| add policy module TeMPLe on top of distillation (ablation) | EgoAdapt · EasyCom ASL | mAP: 78.59 -> 89.74 | Improved |
| use learned policy instead of random policy (ablation) | CFD · EasyCom ASL | mAP: 67.41 -> 89.74 | Improved |
| use learned policy instead of heuristic selection (ablation) | CFD · EasyCom ASL | mAP: 65.88 -> 89.74 | Improved |
| add stage 3 joint finetuning (ablation) | EgoAdapt · EasyCom ASL | mAP: 83.64 -> 89.74 | Improved |
| policy-driven modality selection instead of all modalities (ablation) | EgoAdapt · EasyCom ASL | mAP and GMACs: 88.07 -> 89.74 mAP, 0.148 -> 0.073 GMACs | Improved |
| add visual modality to 4-channel audio (ablation) | fixed fusion · EasyCom ASL | mAP: 81.27 -> 83.44 | Improved |
| learned frame selection instead of random frame (ablation) | EgoAdapt · EPIC-Kitchens | action accuracy: 48.63 -> 56.74 | Improved |
| learned frame selection instead of using all frames (ablation) | EgoAdapt · EPIC-Kitchens | GMACs: 16.51 -> 7.14 GMACs | Improved |
| increase quantization from 4-bit to 32-bit | EgoAdapt · EasyCom ASL | mAP: 78.92 -> 89.74 | Improved |
| swap teacher from TIM to MBT | EgoAdapt · EPIC-Kitchens | action accuracy: 56.74 -> 42.53 | Hurt |
| swap teacher from MUST to GLC | EgoAdapt · AEA behavior anticipation | gaze MAE T700ms: 14.58 -> 19.32 | Hurt |
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.
- AdaMML — AdaMML: Adaptive Multi-Modal Learning for Efficient Video Recognition (ICCV 2021) · 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
- EgoM2P: Egocentric Multimodal Multitask Pretraining (ICCV 2025 · similarity 0.89)
- Ego3S: Select, Strengthen, and Synchronize for Efficient Egocentric Reasoning (ICML 2026 · similarity 0.88)
- Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos (ICLR 2025 · similarity 0.88)
- EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT (NeurIPS 2025 · similarity 0.87)
- EgoGen: An Egocentric Synthetic Data Generator (CVPR 2024 · similarity 0.87)
- EARL: Towards a Unified Analysis-Guided Reinforcement Learning Framework for Egocentric Interaction Reasoning and Pixel Grounding (ICML 2026 · 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.
