Research map / Video / Action recognition
W-TALC: Weakly-supervised Temporal Activity Localization and Classification
Sujoy Paul, Sourya Roy, Amit K. Roy-Chowdhury
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 Co-Activity Similarity Loss (CASL) to MILL (ablation) | W-TALC with I3D features · Thumos14 detection, IoU 0.5 | mAP: 16.9 -> 22.0 | Improved |
| add CASL to MILL (ablation) | W-TALC with I3D features · Thumos14 detection, IoU 0.3 | mAP: 31.2 -> 39.2 | Improved |
| add CASL to MILL (ablation) | W-TALC with I3D features · Thumos14 detection, IoU 0.1 | mAP: 46.5 -> 53.7 | Improved |
| add CASL to MILL (ablation) | W-TALC with I3D features · Thumos14 detection, IoU 0.7 | mAP: 4.4 -> 7.3 | Improved |
| balance MILL and CASL (lambda=0.5) instead of MILL only (lambda=1.0) (ablation) | W-TALC with I3D features · Thumos14 detection | mAP: decrease of 7-8% in mAP with MILL only | Improved |
| use 20-class training only (no extra classes) with I3D features | MILL+CASL+I3DF · Thumos14 detection, IoU 0.5 | mAP: 22.0 -> 22.8 | Improved |
| increase maximum training sequence length T up to 320 s (ablation) | W-TALC · Thumos14 detection (IoU 0.3) | mAP and training time: plateau at T=320 s, training time increases | Mixed |
| use I3D features instead of UntrimmedNet features with CASL | W-TALC · Thumos14 detection, IoU 0.5 | mAP: 18.8 -> 22.0 | Improved |
| add CASL on UntrimmedNet features versus STPN (UNTF) | W-TALC · Thumos14 detection, IoU 0.5 | mAP: 16.2 -> 18.8 | Improved |
Closest papers by meaning
- METAL: Minimum Effort Temporal Activity Localization in Untrimmed Videos (CVPR 2020 · similarity 0.92)
- Weakly Supervised Temporal Action Localization Through Learning Explicit Subspaces for Action and Context (AAAI 2021 · similarity 0.91)
- Adversarial Background-Aware Loss for Weakly-supervised Temporal Activity Localization (ECCV 2020 · similarity 0.91)
- AutoLoc: Weakly-supervised Temporal Action Localization in Untrimmed Videos (ECCV 2018 · similarity 0.91)
- Weakly-Guided Self-Supervised Pretraining for Temporal Activity Detection (AAAI 2023 · similarity 0.90)
- Equivalent Classification Mapping for Weakly Supervised Temporal Action Localization (TPAMI 2022 · similarity 0.90)
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.
