Research map / Label-efficient and robust learning / Knowledge distillation
Learning an Augmented RGB Representation With Cross-Modal Knowledge Distillation for Action Detection
Rui Dai, Srijan Das, François Bremond
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 Atomic KD loss (ablation) | Vanilla-RGB · PKU-MMD CS | mAP: 79.6 -> 82.7 | Improved |
| add Global Contextual Relation loss (ablation) | Vanilla-RGB · PKU-MMD CS | mAP: 79.6 -> 83.7 | Improved |
| add Boundary Saliency loss (ablation) | Vanilla-RGB · PKU-MMD CS | mAP: 79.6 -> 83.1 | Improved |
| add Atomic KD loss (ablation) | Vanilla-RGB · Charades | mAP: 22.3 -> 23.9 | Improved |
| add Boundary Saliency loss to Atomic + Global (ablation) | Atomic + Global · Charades | mAP: 24.4 -> 24.6 | Improved |
| add Boundary Saliency loss to Atomic + Global (ablation) | Atomic + Global · PKU-MMD CS | mAP: 84.3 -> 85.5 | Improved |
| distill at feature level instead of logit level (ablation) | Logit · PKU-MMD CS | mAP: 84.9 -> 85.5 | Improved |
| add logit distillation to feature distillation (ablation) | Feature (Ours) · Charades | mAP: 24.6 -> 24.2 | Hurt |
| use contrastive Atomic KD instead of cosine-distance GD loss | LGD · Charades | mAP: 23.3 -> 23.9 | Improved |
| add 3D pose teacher to OF teacher (ablation) | OF Augmented RGB · PKU-MMD CS | mAP: 85.5 -> 86.3 | Improved |
| use pose teacher instead of OF teacher (ablation) | OF Augmented RGB · TSU-CV | mAP: 19.3 -> 23.6 | Improved |
| apply distillation to TGM temporal filter (ablation) | Vanilla-RGB (TGM) · Charades | mAP: 18.9 -> 21.2 | Improved |
Closest papers by meaning
- Decomposed Cross-Modal Distillation for RGB-Based Temporal Action Detection (CVPR 2023 · similarity 0.94)
- Graph Distillation for Action Detection with Privileged Modalities (ECCV 2018 · similarity 0.90)
- XKD: Cross-Modal Knowledge Distillation with Domain Alignment for Video Representation Learning (AAAI 2024 · similarity 0.89)
- Generative Model-Based Feature Knowledge Distillation for Action Recognition (AAAI 2024 · similarity 0.87)
- Local-Global Multi-Modal Distillation for Weakly-Supervised Temporal Video Grounding (AAAI 2024 · similarity 0.87)
- Video Event Extraction with Multi-View Interaction Knowledge Distillation (AAAI 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.
