Research map / Video / Action recognition
Use Your Head: Improving Long-Tail Video Recognition
Toby Perrett, Saptarshi Sinha, Tilo Burghardt, Majid Mirmehdi, Dima Damen
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 LMR (reconstruction + label mixing) over CE baseline | Motionformer · EPIC-KITCHENS-100 Verbs Val | Avg C/A: 21.2 -> 39.7 | Improved |
| Add LMR over CE baseline | Motionformer · SSv2-LT | Avg C/A: 29.7 -> 38.3 | Improved |
| Add LMR over Framestack on ResNet-50 features | ResNet-50 frame features (VideoLT-LT) · VideoLT-LT | Avg C/A: 41.5 -> 48.9 | Improved |
| Remove residual connection to original features (c(y)=1) (ablation) | LMR · SSv2-LT | Avg C/A: 38.3 -> 6.0 | Hurt |
| Replace learned class contribution with constant 0.5 (ablation) | LMR · EPIC-KITCHENS-100 Verbs Val | Avg C/A: 39.7 -> 39.3 | Hurt |
| Replace learned class contribution with constant 0.5 (ablation) | LMR · SSv2-LT | Avg C/A: 38.3 -> 37.1 | Hurt |
| Remove reconstruction (keep label mixing) (ablation) | LMR · SSv2-LT | Avg C/A: 38.3 -> 37.6 | Hurt |
| Remove reconstruction (keep label mixing) (ablation) | LMR · EPIC-KITCHENS-100 Verbs Val | Avg C/A: 39.7 -> 38.1 | Hurt |
| Remove pairwise label mixing (keep reconstruction) (ablation) | LMR · SSv2-LT | Avg C/A: 38.3 -> 37.9 | Hurt |
| Remove pairwise label mixing (keep reconstruction) (ablation) | LMR · EPIC-KITCHENS-100 Verbs Val | Avg C/A: 39.7 -> 37.1 | Hurt |
| Set few-shot threshold for reconstruction mask to omega=0 (no masking) vs 20 (ablation) | LMR · SSv2-LT | Avg C/A (Acc): 38.3 -> 37.9 | No clear effect |
| Increase samples added per class (+x) on SSv2-LT (ablation) | CE (dataset-level analysis) · SSv2-LT | Few-shot accuracy: maximum gain around +20 samples/class | Improved |
Closest papers by meaning
- MEID: Mixture-of-Experts with Internal Distillation for Long-Tailed Video Recognition (AAAI 2023 · similarity 0.92)
- VideoLT: Large-Scale Long-Tailed Video Recognition (ICCV 2021 · similarity 0.91)
- Bag of Tricks for Long-Tailed Visual Recognition with Deep Convolutional Neural Networks (AAAI 2021 · similarity 0.89)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.88)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.88)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · 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.
