Research map / Vision-language and multimodal / Cross-modal learning
Towards Good Practices for Missing Modality Robust Action Recognition
Sangmin Woo, Sumin Lee, Yeonju Park, Muhammad Adi Nugroho, Changick Kim
This paper tried a common change: Add dropout
Add dropout with p=0.1 · NTU60 · accuracy: 93.3 -> 92.6 Hurt
What 795 accepted papers report for this change
- 549 improved
- 87 no clear effect
- 12 mixed
- 147 hurt
- Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference ICML 2026 · apply layer dropout to whole transformer blocks instead of attention and FFN sub-blocks separately · 20 TPP val loss, 271M · val loss: 2.377 -> 2.367Improved
- Quantifying the Uncertainty of Foundation Models with Singular Value Ensembles ICML 2026 · use MC Dropout · vision and SST-2 · accuracy: 86.3 -> 76.6 on Flowers102Hurt
- Motion Planning in Compressed Representation Spaces ICML 2026 · add nested dropout during training · WOMD guided generation · goal FDE: 4.145 -> 1.104Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 |
|---|---|---|---|
| Use transformer fusion instead of summation | ResNet34 multimodal classifier · NTU60, missing modality settings | mean absolute accuracy discrepancy: 34.8 -> 21.0 | Improved |
| Add ImageNet initialization to the space encoder | ResNet34 multimodal classifier · NTU60, RGB+D+I and RGB+D to RGB | accuracy: 85.9 -> 89.2; 34.8 -> 39.5 | Improved |
| Add shifted center crop | ResNet34 multimodal classifier · NTU60 | accuracy: 89.2 -> 90.5; 39.5 -> 40.3 | Improved |
| Add temporal random sampling | ResNet34 multimodal classifier · NTU60 | accuracy: 90.5 -> 91.0; 40.3 -> 41.4 | Improved |
| Apply color jitter to RGB | ResNet34 multimodal classifier · NTU60 | accuracy: 91.0 -> 92.2; 41.4 -> 42.0 | Improved |
| Add label smoothing with α=0.1 | ResNet34 multimodal classifier · NTU60 | accuracy: 92.2 -> 93.3; 42.0 -> 42.5 | Improved |
| Apply color jitter to depth maps (ablation) | ResNet34 multimodal classifier · NTU60, all modalities | accuracy: 93.3 -> 91.5 | Hurt |
| Add dropout with p=0.1 (ablation) | ResNet34 multimodal classifier · NTU60 | accuracy: 93.3 -> 92.6 | Hurt |
| Add ActionMAE feature reconstruction and randomized modality dropping | Transformer-fusion classifier · NTU60, infrared-only inference | accuracy: 35.4% -> 81.7% | Improved |
| Add ActionMAE to transformer fusion | Transformer-fusion classifier · NTU60, all modalities | accuracy: 92.6% -> 93.0% | Improved |
| Add a memory token (ablation) | ActionMAE (R+D) · NTU60, RGB+D | accuracy: 87.3 -> 92.5 | Improved |
| Add a memory token (ablation) | ActionMAE (R+D) · NTU60, RGB-only inference | accuracy: 74.8 -> 84.5 | Improved |
Closest papers by meaning
- Do More Modalities Always Help? A Geometric Perspective on Missing-Modality Robustness (NeurIPS 2026 · similarity 0.90)
- Quantifying and Enhancing Multi-modal Robustness with Modality Preference (ICLR 2024 · similarity 0.87)
- A Study of Dropout-Induced Modality Bias on Robustness to Missing Video Frames for Audio-Visual Speech Recognition (CVPR 2024 · similarity 0.86)
- Deep Correlated Prompting for Visual Recognition with Missing Modalities (NeurIPS 2024 · similarity 0.86)
- A Closer Look at Multimodal Representation Collapse (ICML 2025 · similarity 0.86)
- Knowledge Bridger: Towards Training-Free Missing Modality Completion (CVPR 2025 · similarity 0.85)
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.
