Summary
Detecting unusual events in videos as they happen is hard without training on specific examples. The authors address this by building a system that looks closely at the current moment in the video to understand what’s happening, then keeps track of relevant past events in a simple way. This avoids heavy computing and large data storage over time. Their approach, called PARSEE-VAD, uses smart reasoning to decide when more detailed analysis is needed and keeps only small, important summaries of past information to spot anomalies quickly. Tests show it works well without needing training and processes video smoothly in real time.
What this means in practice
- •For security monitoring teams: Implement real-time anomaly detection systems that flag unusual video events without requiring prior model training or heavy computation.
- •For autonomous vehicle developers: Enhance onboard video understanding by efficiently detecting anomalies using current scene reasoning and minimal memory of past events.
Authors
Ji Wang, Shuangqing Zhang, Guo-Sen Xie, Fang Zhao
Abstract
Training-free online video anomaly detection (VAD) with frozen multimodal language models faces two coupled challenges: extracting reliable current-window semantics under causal and computational constraints, and maintaining temporal continuity without repeatedly transmitting high-dimensional history. Encoding history through text can compress visual evidence and introduce semantic bias, whereas retaining visual history expands multimodal context. We introduce PARSEE-VAD, a two-module framework that separates semantic evidence acquisition from score-state evolution. Proposition-Aware Reasoning (PAR) extracts structured propositional evidence from the current causal window and conditionally activates more specific queries when coarse evidence warrants further refinement. By sharing a reusable causal visual prefix across queries, PAR reduces redundant computation through selective execution. Streaming Evidence Escalation (SEE) maps the acquired proposition evidence into a compact score-domain event state through current evidence escalation, then propagates only the resulting bounded state across decisions to support temporal continuity. Experiments on four benchmarks demonstrate strong training-free online performance while selective routing reduces specialist computation and score-state propagation remains sparse. These results support a current-first principle for streaming multimodal inference: resolve present semantics first, then use compact historical state only to repair residual continuity gaps.