Research map / Video / Video segmentation and tracking
PiercingEye: Dual-Space Video Violence Detection With Hyperbolic Vision-Language Guidance
Jiaxu Leng, Zhanjie Wu, Mingpi Tan, Mengjingcheng Mo, Jiankang Zheng, Qingqing Li, Ji Gan, Xinbo Gao
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 hyperbolic branch to Euclidean GCN (HE-GCN) (ablation) | PiercingEye · XD-Violence (multimodal) | AP (%): 84.04 -> 86.46 | Improved |
| add hyperbolic branch to Euclidean GCN (HE-GCN) (ablation) | PiercingEye · XD-Violence (unimodal) | AP (%): 77.95 -> 79.70 | Improved |
| add Dual-Space Interaction (DSI) module (ablation) | PiercingEye · XD-Violence (multimodal) | AP (%): 86.46 -> 87.61 | Improved |
| add Dual-Space Interaction (DSI) module (ablation) | PiercingEye · XD-Violence (unimodal) | AP (%): 79.70 -> 82.01 | Improved |
| add HVLGL module (ablation) | PiercingEye · XD-Violence (multimodal) | AP (%): 87.61 -> 88.82 | Improved |
| add HVLGL module (ablation) | PiercingEye · XD-Violence (unimodal) | AP (%): 82.01 -> 83.74 | Improved |
| use hyperbolic-only instead of Euclidean-only features (ablation) | PiercingEye · XD-Violence (multimodal) | AP (%): 84.04 -> 83.67 | Hurt |
| use HE-GCN layer-sensitive node selection instead of HGCN (ablation) | PiercingEye · XD-Violence (multimodal) | AP (%): 84.41 -> 88.82 | Improved |
| use HE-GCN layer-sensitive node selection instead of HGCN (ablation) | PiercingEye · XD-Violence (unimodal) | AP (%): 80.33 -> 83.74 | Improved |
| use Lorentzian metric instead of cosine metric in DSI (ablation) | PiercingEye · XD-Violence (unimodal) | AP (%): 80.50 -> 83.74 | Improved |
| use Lorentzian metric instead of cosine metric in DSI (ablation) | PiercingEye · XD-Violence (multimodal) | AP (%): 86.89 -> 88.82 | Improved |
| use HVLGL instead of Euclidean contrastive loss (ablation) | PiercingEye · XD-Violence (unimodal) | AP (%): 79.43 -> 83.74 | Improved |
Closest papers by meaning
- Beyond Euclidean: Dual-Space Representation Learning for Weakly Supervised Video Violence Detection (NeurIPS 2024 · similarity 0.93)
- Hyperbolic Safety-Aware Vision-Language Models (CVPR 2025 · similarity 0.83)
- Not only Look, but also Listen: Learning Multimodal Violence Detection under Weak Supervision (ECCV 2020 · similarity 0.83)
- TLMA: Mitigating the Impact of Weakly Labeled Information for Video Anomaly Detection (CVPR 2026 · similarity 0.83)
- VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detection (AAAI 2024 · similarity 0.82)
- Learning Event Completeness for Weakly Supervised Video Anomaly Detection (ICML 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.
