Research map / Video / Action recognition
MEID: Mixture-of-Experts with Internal Distillation for Long-Tailed Video Recognition
Xinjie Li, Huijuan Xu
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 |
|---|---|---|---|
| Use MEID instead of FrameStack | VideoLT with ResNet-101 and NetVLAD · VideoLT | Overall mAP: 0.670 -> 0.716 | Improved |
| Use MEID instead of CBS | CharadesLT · CharadesLT | Overall mAP: 0.129 -> 0.151 | Improved |
| Use MEID instead of CBS | CharadesEgoLT · CharadesEgoLT | Overall mAP: 0.163 -> 0.197 | Improved |
| Add frame attention to the first expert (ablation) | First expert on VideoLT · VideoLT | Overall mAP: 0.564 -> 0.580 | Improved |
| Add distillation loss alongside frame attention (ablation) | First expert on VideoLT · VideoLT | Overall mAP: 0.580 -> 0.588 | Improved |
| Add complementary frame selection (ablation) | Second expert on VideoLT · VideoLT | Overall mAP: 0.588 -> 0.604 | Improved |
| Add motion extraction (ablation) | Second expert with CFS on VideoLT · VideoLT | Overall mAP: 0.604 -> 0.615 | Improved |
| Add class-balanced sampling (ablation) | Second expert with CFS and motion extraction on VideoLT · VideoLT | Overall mAP: 0.615 -> 0.619 | Improved |
| Add a designed second expert instead of duplicating the first expert (ablation) | Two-expert MEID on VideoLT · VideoLT | Overall mAP: 0.592 -> 0.619 | Improved |
| Set complementary frame selection β to 0.5 rather than 0.3 (ablation) | MEID on VideoLT · VideoLT | Overall mAP: 0.611 -> 0.619 | Improved |
| Set complementary frame selection β to 0.5 rather than 0.7 (ablation) | MEID on VideoLT · VideoLT | Overall mAP: 0.612 -> 0.619 | Improved |
Closest papers by meaning
- Use Your Head: Improving Long-Tail Video Recognition (CVPR 2023 · similarity 0.92)
- VideoLT: Large-Scale Long-Tailed Video Recognition (ICCV 2021 · similarity 0.91)
- Long-Tailed Multi-Label Visual Recognition by Collaborative Training on Uniform and Re-Balanced Samplings (CVPR 2021 · similarity 0.90)
- Self Supervision to Distillation for Long-Tailed Visual Recognition (ICCV 2021 · similarity 0.90)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.87)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · 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.
