Research map / Architectures and efficiency / Attention and transformers
Two-Stream Convolution Augmented Transformer for Human Activity Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 THAT instead of the strongest baseline, ABLSTM | WiFi-based HAR · Office Room | accuracy: 97.1 -> 98.2 | Helped |
| use THAT instead of the strongest baseline, ABLSTM | WiFi-based HAR · Activity Room | accuracy: 95.6 -> 98.4 | Helped |
| use THAT instead of the strongest baseline, ABLSTM | WiFi-based HAR · Meeting Room | accuracy: 96.8 -> 99.0 | Helped |
| use THAT instead of the strongest baseline, ABLSTM | WiFi-based HAR · Activity+Meeting | accuracy: 95.9 -> 98.6 | Helped |
| add Gaussian Range Encoding (ablation) | THAT · four evaluation datasets | accuracy: 97.9 -> 98.2; 97.9 -> 98.4; 98.5 -> 99.0; 98.2 -> 98.6 | Helped |
| use Gaussian Range Encoding instead of positional encoding (ablation) | THAT · four evaluation datasets | accuracy: 91.1 -> 98.2; 84.3 -> 98.4; 90.3 -> 99.0; 87.8 -> 98.6 | Helped |
| add Multi-scale CNN (ablation) | THAT · four evaluation datasets | accuracy: 95.3 -> 98.2; 97.4 -> 98.4; 97.0 -> 99.0; 97.3 -> 98.6 | Helped |
| use Multi-scale CNN instead of PFFN (ablation) | THAT · four evaluation datasets | accuracy: 97.2 -> 98.2; 97.2 -> 98.4; 98.1 -> 99.0; 97.7 -> 98.6 | Helped |
| add Temporal Module (ablation) | THAT · four evaluation datasets | accuracy: 92.0 -> 98.2; 95.2 -> 98.4; 97.5 -> 99.0; 95.9 -> 98.6 | Helped |
| add Channel Module (ablation) | THAT · four evaluation datasets | accuracy: 93.8 -> 98.2; 97.7 -> 98.4; 98.3 -> 99.0; 98.0 -> 98.6 | Helped |
| use THAT instead of ABLSTM | WiFi-based HAR on Activity Room · Activity Room | testing speed: 1.83× faster | Helped |
Closest papers by meaning
- ConSense: Continually Sensing Human Activity with WiFi via Growing and Picking (AAAI 2025 · similarity 0.89)
- WiFi CSI Based Temporal Activity Detection via Dual Pyramid Network (AAAI 2025 · similarity 0.88)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.87)
- CSI-Bench: A Large-Scale In-the-Wild Dataset for Multi-task WiFi Sensing (NeurIPS 2025 · similarity 0.85)
- Scalable Hierarchical Deep Neural Network for Time Series Analysis in Wearable Sensor-based Human Activity Recognition (AAAI 2025 · similarity 0.84)
- MoPFormer: Motion-Primitive Transformer for Wearable-Sensor Activity Recognition (NeurIPS 2025 · similarity 0.82)
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.