Research map / Video / Action recognition
Event Stream based Human Action Recognition: A High-Definition Benchmark Dataset and Algorithms
Xiao Wang, Shiao Wang, Pengpeng Shao, Lin Zhu, Bo Jiang, Yonghong Tian
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 voxel temporal scanning (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.720 -> 0.723 | Improved |
| add voxel temporal scanning (ablation) | EVMamba · Bully10K | top-1 accuracy: 91.17 -> 91.52 | Improved |
| add voxel temporal scanning (ablation) | EVMamba · DailyDVS-200 | top-1 accuracy: 49.23 -> 49.65 | Improved |
| add event frames to voxel-only input (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.451 -> 0.723 | Improved |
| add event voxel to event frame input (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.720 -> 0.723 | Improved |
| use additive fusion instead of concatenation (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.717 -> 0.723 | Improved |
| use additive fusion instead of A-matrix interaction (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.716 -> 0.723 | Improved |
| use additive fusion instead of B-matrix interaction (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.717 -> 0.723 | Improved |
| use trajectory threshold 1 instead of 9 (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.712 -> 0.723 | Improved |
| use 8 input frames instead of 4 (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.698 -> 0.723 | Improved |
| use 8 input frames instead of 12 (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.708 -> 0.723 | Improved |
| use 224x224 input instead of 192x192 (ablation) | EVMamba · CeleX-HAR | top-1 accuracy: 0.704 -> 0.723 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ACTION-Net — ACTION-Net: Multipath Excitation for Action Recognition (CVPR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.91)
- DarkShake-DVS: Event-based Human Action Recognition under Low-light and Shaking Camera Conditions (CVPR 2026 · similarity 0.91)
- Action Recognition and Benchmark Using Event Cameras (TPAMI 2023 · similarity 0.90)
- Human Action Recognition From Various Data Modalities: A Review (TPAMI 2022 · similarity 0.90)
- Event Stream-based Visual Object Tracking: A High-Resolution Benchmark Dataset and A Novel Baseline (CVPR 2024 · similarity 0.87)
- PASS: Path-selective State Space Model for Event-based Recognition (NeurIPS 2025 · similarity 0.87)
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.
