Research map / Video / Action recognition
Deep Bilinear Learning for RGB-D Action Recognition
Jian-Fang Hu, Wei-Shi Zheng, Jiahui Pan, 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 |
|---|---|---|---|
| add temporal modelling over action history subsequences (AHS) (ablation) | Deep Bilinear RGB+DEP+SKL · NTU RGB+D cross-subject | accuracy (%): 83.5% -> 85.4% | Improved |
| add temporal modelling (RGB only) (ablation) | Deep Bilinear RGB · NTU RGB+D cross-subject | accuracy (%): 72.5% -> 75.3% | Improved |
| add temporal modelling (skeleton only) (ablation) | Deep Bilinear SKL · NTU RGB+D cross-subject | accuracy (%): 69.8% -> 72.5% | Improved |
| include only AHSs of length at least 7 instead of at least 1 (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 84.1% -> 85.4% | Improved |
| also include very short AHSs (>=1 versus >=7) (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 85.4% -> 84.1% | Hurt |
| use 3 bilinear blocks instead of 1 (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 83.8% -> 85.4% | Improved |
| increase bilinear blocks beyond 3 (3 to 5) (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 85.4% -> 84.9% | No clear effect |
| replace plane-wise pooling with element-wise FCN (ablation) | Deep Bilinear bilinear block · NTU RGB+D cross-subject | accuracy (%): 85.4% -> 76.3% | Hurt |
| use learned bilinear pooling instead of mean pooling (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 83.0% -> 85.4% | Improved |
| use learned bilinear pooling instead of max pooling (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 77.5% -> 85.4% | Improved |
| use proposed bilinear blocks instead of MCB (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 84% -> 85.4% | Improved |
| fuse temporal dimension first, then modality, in each block (ablation) | Deep Bilinear · NTU RGB+D cross-subject | accuracy (%): 85.4% -> 85.0% | No clear effect |
Closest papers by meaning
- Deep Multimodal Feature Analysis for Action Recognition in RGB+D Videos (TPAMI 2017 · similarity 0.88)
- Temporal Bilinear Networks for Video Action Recognition (AAAI 2019 · similarity 0.87)
- Cooperative Training of Deep Aggregation Networks for RGB-D Action Recognition (AAAI 2018 · similarity 0.87)
- Modality Distillation with Multiple Stream Networks for Action Recognition (ECCV 2018 · similarity 0.86)
- Decomposed Cross-Modal Distillation for RGB-Based Temporal Action Detection (CVPR 2023 · similarity 0.86)
- MMNet: A Model-Based Multimodal Network for Human Action Recognition in RGB-D Videos (TPAMI 2022 · 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.
