Research map / Video / Action recognition
Two-Stream Consensus Network for Weakly-Supervised Temporal Action Localization
Yuanhao Zhai, Le Wang, Wei Tang, Qilin Zhang, Junsong Yuan, Gang Hua
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 hard pseudo ground truth from fused attention (fusion) (ablation) | TSCN with UntrimmedNet · THUMOS14 test | mAP@IoU 0.5: 24.0 -> 27.6 | Improved |
| add hard pseudo ground truth to RGB stream (ablation) | RGB stream · THUMOS14 test | mAP@IoU 0.5: 8.2 -> 14.4 | Improved |
| add hard pseudo ground truth to flow stream (ablation) | flow stream · THUMOS14 test | mAP@IoU 0.5: 23.2 -> 24.1 | Improved |
| add hard pseudo ground truth to flow stream (recall) (ablation) | flow stream · THUMOS14 test | recall at IoU 0.5: 43.3 -> 42.4 | Hurt |
| add hard pseudo ground truth to fusion (precision) (ablation) | TSCN fusion · THUMOS14 test | precision at IoU 0.5: 23.6 -> 31.3 | Improved |
| use hard instead of soft pseudo ground truth (ablation) | TSCN · THUMOS14 test | mAP | Improved |
| add attention normalization loss L_att instead of background classification loss L_bg (ablation) | TSCN · THUMOS14 test | mAP | Improved |
| add background classification loss on top of L_att (ablation) | TSCN · THUMOS14 test | mAP | Hurt |
| switch features from UntrimmedNet to I3D | TSCN · THUMOS14 test | mAP@IoU 0.5: 27.6 -> 28.7 | Improved |
| TSCN versus BaS-Net with I3D features at high IoU | TSCN · THUMOS14 test | mAP@IoU 0.7: 10.4 -> 10.2 | Hurt |
| TSCN versus Nguyen et al. on ActivityNet v1.3 at IoU 0.5 | TSCN · ActivityNet v1.3 val | mAP@IoU 0.5: 36.4 -> 35.3 | Hurt |
| TSCN versus Liu et al. on ActivityNet v1.2 | TSCN · ActivityNet v1.2 val | average mAP: 22.4 -> 23.6 | Improved |
Closest papers by meaning
- Adaptive Two-Stream Consensus Network for Weakly-Supervised Temporal Action Localization (TPAMI 2022 · similarity 0.99)
- Two-Stream Networks for Weakly-Supervised Temporal Action Localization With Semantic-Aware Mechanisms (CVPR 2023 · similarity 0.92)
- Weakly Supervised Temporal Action Localization Through Contrast Based Evaluation Networks (ICCV 2019 · similarity 0.91)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.89)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.89)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.89)
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.
