Research map / Video / Action recognition
Temporal Recurrent Networks for Online Action Detection
Mingze Xu, Mingfei Gao, Yi-Ting Chen, Larry S. Davis, David J. Crandall
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 predicted future context (TRN cell) to LSTM | LSTM · HDD (Multimodal input) | mAP: 32.7 -> 40.8 | Improved |
| add predicted future context to LSTM | LSTM · TVSeries (VGG) | mcAP: 64.1 -> 75.4 | Improved |
| add predicted future context to LSTM | LSTM · THUMOS'14 | mAP: 39.3 -> 47.2 | Improved |
| use TRN future gate instead of plain encoder-decoder (ablation) | ED · HDD (Multimodal input) | mAP: 37.8 -> 40.8 | Improved |
| feed true future frame features instead of predicted ones (oracle RNN-offline) | RNN · HDD | mAP: 32.7 -> 41.6 | Improved |
| use predicted future (TRN) instead of true future (RNN-offline) | TRN · THUMOS'14 | mAP: 47.3 -> 47.2 | No clear effect |
| add optical flow (two-stream features) to VGG features | TRN · TVSeries | mcAP: 75.4 -> 83.7 | Improved |
| add optical flow motion input to single-frame CNN | CNN · THUMOS'14 | mAP: 34.7 -> 36.2 | Improved |
| increase decoder steps from 4 to 8 (ablation) | TRN · THUMOS'14 | mAP: 46.0 -> 47.2 | Mixed |
| increase decoder steps from 4 to 10 (ablation) | TRN · HDD | action anticipation mAP: 34.3 -> 25.4 | Hurt |
| increase decoder steps from 4 to 10 (ablation) | TRN · THUMOS'14 | action anticipation mAP: 42.6 -> 35.0 | Hurt |
| TRN vs ED/RED on anticipation | TRN · THUMOS'14 (anticipation) | mAP: 37.5 -> 38.9 | Improved |
Closest papers by meaning
- End-To-End Learning of Action Detection From Frame Glimpses in Videos (CVPR 2016 · similarity 0.89)
- Temporal Segment Networks for Action Recognition in Videos (TPAMI 2018 · similarity 0.88)
- TDN: Temporal Difference Networks for Efficient Action Recognition (CVPR 2021 · similarity 0.88)
- Recurrent Tubelet Proposal and Recognition Networks for Action Detection (ECCV 2018 · similarity 0.88)
- Online Human Action Detection Using Joint Classification-Regression Recurrent Neural Networks (ECCV 2016 · similarity 0.88)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 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.
