Research map / Video / Action recognition
VSViG: Real-time Video-based Seizure Detection via Skeleton-based Spatiotemporal ViG
Yankun Xu, Junzhe Wang, Yun-Hsuan Chen, Jie Yang, Wenjie Ming, Shuang Wang, Mohamad Sawan
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 both inter-partition and intra-partition spatial modules (ablation) | VSViG · patient video validation | RMSE: 13.0 -> 5.9 | Improved |
| add intra-partition module on top of inter-only (ablation) | VSViG · patient video validation | RMSE: 11.7 -> 5.9 | Improved |
| use K_S=3 spatial neighbors in temporal convolution instead of K_S=1 (ablation) | VSViG · patient video validation | RMSE: 9.9 -> 5.9 | Improved |
| use temporal kernel K_T=5 instead of 3 (ablation) | VSViG · patient video validation | RMSE: 5.9 -> 6.5 | Hurt |
| add dynamic partition (shuffle partitions before temporal update) (ablation) | VSViG · patient video validation | RMSE: 8.4 -> 5.9 | Improved |
| add Stem(x, y) positional embedding (ablation) | VSViG · patient video validation | RMSE: 9.1 -> 5.9 | Improved |
| use Gaussian sigma 0.3 rather than 0.1 for patch extraction (ablation) | VSViG · patient video validation | RMSE: 10.1 -> 5.9 | Improved |
| use exponential rather than linear transition-period labels (ablation) | VSViG · patient video validation | RMSE: 6.7% -> 5.9% | Improved |
| add accumulative probability decision rule (ablation) | VSViG · seizure onset detection DT=0.3 | latency after EEG onset: 11.4s -> 5.1s | Improved |
| use skeleton-based skeleton models instead of RGB-based models | seizure detection baselines · patient video validation | RMSE (%): R3D+ViT 14.2 -> HDGCN 7.9 | Improved |
| use skeleton-based patch embeddings instead of coordinates (VSViG vs best GCN) | skeleton action recognition · patient video validation | RMSE (%): 7.9 -> 5.9 | Improved |
Closest papers by meaning
- Long-Term EEG Partitioning for Seizure Onset Detection (AAAI 2025 · similarity 0.86)
- EvoBrain: Dynamic Multi-Channel EEG Graph Modeling for Time-Evolving Brain Networks (NeurIPS 2025 · similarity 0.86)
- DMNet: Self-comparison Driven Model for Subject-independent Seizure Detection (NeurIPS 2024 · similarity 0.86)
- REST: Efficient and Accelerated EEG Seizure Analysis through Residual State Updates (ICML 2024 · similarity 0.85)
- Seizure-Semiology-Suite($S^3$): A Clinically Multimodal Dataset, Benchmark, and Models for Seizure Semiology Understanding (ICML 2026 · similarity 0.84)
- FAPEX: Fractional Amplitude-Phase Expressor for Robust Cross-Subject Seizure Prediction (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.
