Research map / Vision-language and multimodal / Cross-modal learning
MMG-Ego4D: Multimodal Generalization in Egocentric Action Recognition
Xinyu Gong, Sreyas Mohan, Naina Dhingra, Jean-Charles Bazin, Yilei Li, Zhangyang Wang, Rakesh Ranjan
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 cross-modal contrastive alignment loss (ablation) | MMG-Ego4D · MMG-Ego4D supervised zero-shot cross-modal, train audio+IMU, test video | Top-1 accuracy: 2.37 -> 25.03 | Improved |
| Add cross-modal contrastive alignment loss (ablation) | MMG-Ego4D · MMG-Ego4D few-shot zero-shot cross-modal, support audio+IMU, query video | 5-way 5-shot accuracy: 33.93 -> 51.40 | Improved |
| Add cross-modal contrastive alignment loss (ablation) | MMG-Ego4D · MMG-Ego4D supervised regular | Top-1 accuracy: 52.18 -> 55.66 | Improved |
| Use attention-based Transformer fusion instead of MLP fusion (ablation) | MMG-Ego4D · MMG-Ego4D supervised regular | Top-1 accuracy: 52.79 -> 55.66 | Improved |
| Use attention-based Transformer fusion instead of MLP fusion (ablation) | MMG-Ego4D · MMG-Ego4D supervised missing modality, test audio+IMU | Top-1 accuracy: 32.89 -> 37.07 | Improved |
| Use attention-based Transformer fusion instead of MLP fusion (ablation) | MMG-Ego4D · MMG-Ego4D few-shot zero-shot cross-modal, support audio+IMU, query video | 5-way 5-shot accuracy: 51.08 -> 51.40 | Improved |
| Add cross-modal prototypical loss (ablation) | MMG-Ego4D · MMG-Ego4D few-shot regular | 5-way 5-shot accuracy: 62.37 -> 63.00 | Improved |
| Add cross-modal prototypical loss (ablation) | MMG-Ego4D · MMG-Ego4D few-shot missing modality, query audio+IMU | 5-way 5-shot accuracy: 50.03 -> 50.77 | Improved |
| Add cross-modal prototypical loss (ablation) | MMG-Ego4D · MMG-Ego4D few-shot zero-shot cross-modal, support audio+IMU, query video | 5-way 5-shot accuracy: 50.80 -> 51.40 | Improved |
| Use modality dropout during fusion training (ablation) | MMG-Ego4D · MMG-Ego4D generalization tasks | accuracy | No clear effect |
| Average all fusion output tokens rather than use the CLS token (ablation) | MMG-Ego4D · MMG-Ego4D action recognition | accuracy | Improved |
| Add modality-specific embeddings before fusion (ablation) | MMG-Ego4D · MMG-Ego4D action recognition | accuracy | Improved |
Closest papers by meaning
- Are We Making Progress in Multimodal Domain Generalization? A Comprehensive Benchmark Study (NeurIPS 2026 · similarity 0.88)
- X-MIC: Cross-Modal Instance Conditioning for Egocentric Action Generalization (CVPR 2024 · similarity 0.87)
- 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)
- Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives (CVPR 2024 · similarity 0.86)
- MMEgo: Towards Building Egocentric Multimodal LLMs for Video QA (ICLR 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.
