Research map / Video / Action recognition
Early Action Prediction by Soft Regression
Jian–Fang Hu, Wei‐Shi Zheng, Lianyang Ma, Gang Wang, Jianhuang Lai, Jianguo Zhang
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 |
|---|---|---|---|
| use learned soft labels instead of random soft labels (ablation) | MSRNN · SYSU 3D HOI | AUC: 0.66871 -> 0.71615 | Improved |
| use learned soft labels instead of all-ones labels (ablation) | MSRNN · SYSU 3D HOI | AUC: 0.68521 -> 0.71615 | Improved |
| use learned soft labels instead of manually increasing 0.5 to 1 (ablation) | MSRNN · SYSU 3D HOI | AUC: 0.70083 -> 0.71615 | Improved |
| add consistency term with weight 500 instead of 0 (ablation) | MSRNN · SYSU 3D HOI | AUC: 0.67031 -> 0.71615 | Improved |
| raise consistency weight to 500000 (ablation) | MSRNN · SYSU 3D HOI | AUC: 0.71615 -> 0.65521 | Hurt |
| use increasing weights s instead of decreasing (ablation) | MSRNN · SYSU 3D HOI | AUC: 0.6875 -> 0.71615 | Improved |
| learn class-specific multiple soft labels instead of a single soft label vector (ablation) | SRNN · SYSU 3D HOI | AUC: 70.2 -> 71.6 | Improved |
| learn class-specific multiple soft labels in linear model (ablation) | SLR · SYSU 3D HOI | AUC: 69.6 -> 70.1 | Improved |
| add depth and skeleton to RGB input (ablation) | MSRNN · SYSU 3D HOI | AUC: 61.3 -> 71.6 | Improved |
| use RGB-D input instead of RGB-only (ablation) | MSRNN · SYSU 3D HOI | AUC: 62.7 -> 71.6 | Improved |
| use CNN features instead of HOG-based LAFF | MSRNN · SYSU 3D HOI | AUC: 71.6 -> 75.4 | Improved |
| use soft regression on two-stream CNN features instead of direct two-stream prediction | MSRNN · SYSU 3D HOI | AUC: 64.9 -> 66.2 | Improved |
Closest papers by meaning
- Action Anticipation with RBF Kernelized Feature Mapping RNN (ECCV 2018 · similarity 0.87)
- Encouraging LSTMs to Anticipate Actions Very Early (ICCV 2017 · similarity 0.86)
- Weakly-Supervised Action Transition Learning for Stochastic Human Motion Prediction (CVPR 2022 · similarity 0.85)
- Modality Distillation with Multiple Stream Networks for Action Recognition (ECCV 2018 · similarity 0.85)
- Scene Flow to Action Map: A New Representation for RGB-D Based Action Recognition With Convolutional Neural Networks (CVPR 2017 · similarity 0.85)
- Weakly Supervised Action Learning With RNN Based Fine-To-Coarse Modeling (CVPR 2017 · similarity 0.85)
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.
