Research map / Video / Action recognition
Weakly Supervised Temporal Action Localization Through Contrast Based Evaluation Networks
Ziyi Liu, Le Wang, Qilin Zhang, Zhanning Gao, Zhenxing Niu, Nanning Zheng, 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 contrast-score proposal evaluator (no trained generator) over thresholding (ablation) | UntrimmedNet · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 15.4 -> 21.6 | Improved |
| add training of action proposal generator (C1) (ablation) | Plain-Model · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 21.6 -> 22.9 | Improved |
| add start/end edge scores to action score in contrast score (ablation) | Actioness-Only · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 1.2 -> 22.9 | Improved |
| add action score to edge scores in contrast score (ablation) | Edgeness-Only · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 11.4 -> 22.9 | Improved |
| add TSN sampling (C4) (ablation) | CleanNet-Simple · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 22.9 -> 23.4 | Improved |
| add joint finetuning of action classification (C5) (ablation) | CleanNet-Simple · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 22.9 -> 23.6 | Improved |
| add joint finetuning on top of TSN sampling (ablation) | CleanNet-T · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 23.4 -> 23.9 | Improved |
| add TSN sampling on top of joint finetuning (ablation) | CleanNet-J · THUMOS14 test (IoU 0.5, ablation) | mAP (%): 23.6 -> 23.9 | Improved |
| CleanNet vs AutoLoc at IoU 0.5 (proposal evaluation differs) | AutoLoc · THUMOS14 test (IoU 0.5) | mAP (%): 21.2 -> 23.9 | Improved |
| CleanNet vs W-TALC on ActivityNet v1.2 at IoU 0.5 (low-noise videos) | W-TALC · ActivityNet v1.2 val (IoU 0.5) | mAP (%): 37.0 -> 37.1 | No clear effect |
| CleanNet vs AutoLoc on ActivityNet v1.2 average mAP | AutoLoc · ActivityNet v1.2 val (avg IoU 0.5:0.95) | mAP (%): 16.0 -> 21.6 | Improved |
Closest papers by meaning
- Fine-Grained Temporal Contrastive Learning for Weakly-Supervised Temporal Action Localization (CVPR 2022 · similarity 0.94)
- Exploring Denoised Cross-Video Contrast for Weakly-Supervised Temporal Action Localization (CVPR 2022 · similarity 0.93)
- ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization (AAAI 2021 · similarity 0.93)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.92)
- Weakly Supervised Temporal Action Localization Through Learning Explicit Subspaces for Action and Context (AAAI 2021 · similarity 0.92)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.91)
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.
