Research map / Detection and segmentation / Open-set and few-shot detection
Graph Distillation for Action Detection with Privileged Modalities
Zelun Luo, Jun-Ting Hsieh, Lu Jiang, Juan Carlos Niebles, Li Fei-Fei
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 a learned distillation graph instead of an empty graph (ablation) | graph distillation · mini-NTU RGB+D | mAP: 0.464 -> 0.619 | Improved |
| use a learned graph instead of a uniform graph (ablation) | graph distillation · mini-NTU RGB+D | mAP: 0.537 -> 0.619 | Improved |
| use a learned graph instead of a prior graph (ablation) | graph distillation · mini-NTU RGB+D | mAP: 0.571 -> 0.619 | Improved |
| use a learned graph instead of a prior graph (ablation) | graph distillation · mini-PKU-MMD (tIoU 0.5) | mAP: 0.515 -> 0.559 | Improved |
| use graph distillation instead of knowledge distillation | knowledge distillation · mini-NTU RGB+D | mAP: 0.524 -> 0.619 | Improved |
| use multi-task privileged learning instead of independent training | empty graph · mini-NTU RGB+D | mAP: 0.464 -> 0.456 | Hurt |
| add privileged modalities in the source domain (ablation) | src + trg · mini-PKU-MMD (tIoU 0.5) | mAP: 0.501 -> 0.533 | Improved |
| add privileged modalities in the target domain on top of source PIs (ablation) | src w/ PIs + trg · mini-PKU-MMD (tIoU 0.5) | mAP: 0.533 -> 0.562 | Improved |
| add source-domain pre-training (transfer from NTU RGB+D) (ablation) | trg only · mini-PKU-MMD (tIoU 0.5) | mAP: 0.200 -> 0.501 | Improved |
| increase target privileged modalities from 1 to all (ablation) | graph distillation · mini-PKU-MMD (tIoU 0.5) | mAP: 0.549 -> 0.562 | Improved |
| add transfer pre-training from NTU RGB+D on full PKU-MMD (ablation) | graph distillation · PKU-MMD, test RGB (tIoU 0.5) | mAP: 0.743 -> 0.801 | Improved |
| distill optical flow into RGB on UCF-101 instead of training from scratch | RGB stream · UCF-101 | mAP: 0.484 -> 0.757 | Improved |
Closest papers by meaning
- Modality Distillation with Multiple Stream Networks for Action Recognition (ECCV 2018 · similarity 0.91)
- Learning an Augmented RGB Representation With Cross-Modal Knowledge Distillation for Action Detection (ICCV 2021 · similarity 0.90)
- Local-Global Multi-Modal Distillation for Weakly-Supervised Temporal Video Grounding (AAAI 2024 · similarity 0.86)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.85)
- A Multimodal, Multi-Task Adapting Framework for Video Action Recognition (AAAI 2024 · similarity 0.85)
- Mosaic of Modalities: A Comprehensive Benchmark for Multimodal Graph Learning (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.
