Research map / Video / Action recognition
RhyRNN: Rhythmic RNN for Recognizing Events in Long and Complex Videos
Tianshu Yu, Yikang Li, Baoxin Li
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 BR to RhyRNN+DivPool (ablation) | RhyRNN · Breakfast event recognition | accuracy (%): 43.7 -> 44.3 | Improved |
| add BR to RhyRNN+DivPool (ablation) | RhyRNN · Breakfast event recognition | mAP (%): 40.5 -> 41.9 | Improved |
| add DivPool to RhyRNN+BR (ablation) | RhyRNN · Breakfast event recognition | accuracy (%): 40.7 -> 44.3 | Improved |
| add DivPool to RhyRNN+BR (ablation) | RhyRNN · Breakfast event recognition | mAP (%): 35.2 -> 41.9 | Improved |
| add DivPool and BR on top of RhyRNN alone (ablation) | RhyRNN · Breakfast event recognition | accuracy (%): 35.8 -> 44.3 | Improved |
| add BR to RhyRNN (ablation) | RhyRNN on I3D features · Charades multi-label recognition | mAP (%): 24.6 -> 25.4 | Improved |
| replace 6-layer IndRNN with 2-layer RhyRNN | sequence model without DivPool/BR · Breakfast event recognition | accuracy (%): 19.4 -> 35.8 | Improved |
| replace SkipRNN with RhyRNN | sequence model without DivPool/BR · Breakfast event recognition | accuracy (%): 31.9 -> 35.8 | Improved |
| replace IndRNN with RhyRNN in DivPool+BR pipeline | DivPool+BR pipeline · Breakfast event recognition | accuracy (%): 42.7 -> 44.3 | Improved |
| replace SkipRNN with RhyRNN in DivPool+BR pipeline | DivPool+BR pipeline · Breakfast event recognition | accuracy (%): 40.2 -> 44.3 | Improved |
| stack 4 RhyRNN layers instead of 2 before one DivPool (ablation) | RhyRNN+DivPool · Breakfast event recognition | accuracy (%): 43.7 -> 43.1 | Hurt |
| use two DivPool stages with 4 RhyRNN layers instead of one with 2 (ablation) | RhyRNN+DivPool · Breakfast event recognition | accuracy (%): 43.7 -> 41.4 | Hurt |
Closest papers by meaning
- CERN: Confidence-Energy Recurrent Network for Group Activity Recognition (CVPR 2017 · similarity 0.86)
- Two-Stream Region Convolutional 3D Network for Temporal Activity Detection (TPAMI 2019 · similarity 0.85)
- Learning Latent Super-Events to Detect Multiple Activities in Videos (CVPR 2018 · similarity 0.85)
- Every Moment Counts: Dense Detailed Labeling of Actions in Complex Videos (IJCV 2017 · similarity 0.85)
- Joint Prediction of Activity Labels and Starting Times in Untrimmed Videos (ICCV 2017 · similarity 0.85)
- ZSTAD: Zero-Shot Temporal Activity Detection (CVPR 2020 · similarity 0.84)
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.
