Research map / Video / Action recognition
WOAD: Weakly Supervised Online Action Detection in Untrimmed Videos
Check 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 WOAD's TPG-generated pseudo frame labels and OAR instead of W-TALC frame scores | Weakly supervised online action detection · THUMOS'14 | mean F-AP: 48.0 -> 54.4 | Helped |
| Use WOAD's TPG-generated pseudo frame labels and OAR instead of W-TALC frame scores | Weakly supervised online action detection · ActivityNet1.2 | mean F-AP: 53.8 -> 66.7 | Helped |
| Use WOAD rather than W-TALC | Weakly supervised action-start detection · THUMOS'14 | mean P-AP@1 second: 16.2 -> 21.9 | Helped |
| Use WOAD rather than W-TALC | Weakly supervised action-start detection · ActivityNet1.2 | mean P-AP@1 second: 5.2 -> 7.9 | Helped |
| Add segment-level annotations to a fraction of videos | WOAD · THUMOS'14 | mean F-AP: 54.4 -> 55.0 -> 59.3 -> 62.6 -> 67.1 | Helped |
| Add segment-level annotations to a fraction of videos | WOAD · ActivityNet1.2 | mean F-AP: 66.7 -> 66.9 -> 68.5 -> 69.3 -> 70.7 | Helped |
| Add TPG loss during joint training (ablation) | Strongly supervised WOAD · THUMOS'14 | mean F-AP: 61.2 -> 67.1 | Helped |
| Add pseudo-label supervision alongside strong labels (ablation) | Strongly supervised WOAD · THUMOS'14 | mean F-AP: 63.9 -> 67.1 | Helped |
| Add start prediction scores at inference (ablation) | WOAD · THUMOS'14 | mean P-AP@1: 20.2 -> 21.9 | Helped |
| Add temporal pooling (ablation) | Strongly supervised WOAD · THUMOS'14 | mean P-AP@1: 26.3 -> 28.0 | Helped |
| Use an LSTM instead of two fully connected layers (ablation) | Weakly supervised WOAD · THUMOS'14 | mean F-AP: 49.0 -> 54.4 | Helped |
| Vary the TPG-loss weight lambda (ablation) | Weakly supervised WOAD · THUMOS'14 | mean F-AP: 54.4, 55.0, 54.6 | No clear effect |
Closest papers by meaning
- Large-Scale Weakly-Supervised Pre-Training for Video Action Recognition (CVPR 2019 · similarity 0.89)
- Does Video-Text Pretraining Help Open-Vocabulary Online Action Detection? (NeurIPS 2024 · similarity 0.86)
- Action-Agnostic Point-Level Supervision for Temporal Action Detection (AAAI 2025 · similarity 0.86)
- Context-Enhanced Memory-Refined Transformer for Online Action Detection (CVPR 2025 · similarity 0.86)
- Stable Mean Teacher for Semi-supervised Video Action Detection (AAAI 2025 · similarity 0.86)
- Semi-supervised Active Learning for Video Action Detection (AAAI 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.