Research map / Video / Action recognition
Weakly-Guided Self-Supervised Pretraining for Temporal Activity Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 weak-label-guided detection pretraining and ensemble augmentations | X3D · Charades | mAP (%): 20.66 -> 23.94 | Helped |
| use weak-label-guided detection pretraining and ensemble augmentations | X3D · MultiTHUMOS | mAP (%): 37.17 -> 40.88 | Helped |
| add Volume Freeze (ablation) | X3D · Charades | mAP (%): 17.28 -> 18.79 | Helped |
| add Volume MixUp (ablation) | X3D · Charades | mAP (%): 17.28 -> 19.18 | Helped |
| add Volume CutMix (ablation) | X3D · Charades | mAP (%): 17.28 -> 18.99 | Helped |
| ensemble separately trained augmentation models instead of jointly applying multiple augmentations (ablation) | X3D · Charades | mAP (%): 19.11 -> 20.50 | Helped |
| apply one randomly selected augmentation per clip rather than multiple augmentations (ablation) | X3D joint training · Charades | mAP (%): 19.11 -> 19.28 | Helped |
| add a second frozen segment (ablation) | Volume Freeze on X3D · Charades | mAP (%): 18.79 -> 18.83 | No clear effect |
| use seamless rather than hard boundaries between mixed clips (ablation) | Volume MixUp on X3D · Charades | mAP (%): 18.86 -> 19.18 | Helped |
| apply MixUp at a random feature-map depth rather than at the input (ablation) | Volume MixUp on X3D · Charades | mAP (%): 19.18 -> 19.24 | No clear effect |
| use transient view rather than transient window (ablation) | Volume CutMix on X3D · Charades | mAP (%): 18.89 -> 18.99 | Helped |
| use Volume CutMix in both Coarse and Fine streams (ablation) | Coarse-Fine · Charades | mAP (%): 23.29 -> 22.83 | Hurt |
Closest papers by meaning
- Activity Driven Weakly Supervised Object Detection (CVPR 2019 · similarity 0.90)
- Learning Activity Progression in LSTMs for Activity Detection and Early Detection (CVPR 2016 · similarity 0.88)
- Fast Temporal Activity Proposals for Efficient Detection of Human Actions in Untrimmed Videos (CVPR 2016 · similarity 0.88)
- Detector-Free Weakly Supervised Group Activity Recognition (CVPR 2022 · similarity 0.88)
- Rethinking Pseudo-Label Guided Learning for Weakly Supervised Temporal Action Localization from the Perspective of Noise Correction (AAAI 2025 · similarity 0.86)
- Learning to Predict Activity Progress by Self-Supervised Video Alignment (CVPR 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.