Decision readout styles influence video anomaly detection accuracy differently
Decision Readouts for Text-Mediated Video Anomaly Detection: An Exploratory Evaluation of Jev and Qwen
Artificial IntelligenceComputer Vision and Pattern Recognition
Summary
Detecting unusual events in videos can be helped by turning video cues into text and then deciding if something is wrong. This paper looks at how different ways of reading and interpreting these text clues affect the detection quality when the underlying video information is the same. The authors tested several methods on crime and violence video sets and found that some readout styles worked better on one dataset but not on another. These results suggest that the way decisions are presented or processed matters, but more research is needed to confirm and understand why.
What this means in practice
- •For security system developers: Improve video anomaly detection accuracy by choosing decision readout styles tailored to specific video datasets.
- •For surveillance software engineers: Refine alert decision interfaces using evaluated readout methods for better precision in recognizing unusual events from video summaries.
Authors
Xukui Qin, Youting Wang, Xinjie He, Ziyang Luo, Runxiong Wu, Yan-Syuan Chen, Zhongyao Chu
Abstract
How much does the decision readout matter when video-derived textual evidence is held fixed? We evaluate Jev typed decisions and three Qwen readouts on a sparse development sample of 40 videos and 400 target anchors from UCF-Crime and XD-Violence, each presented as a summary and ordered captions. Each dataset contributes 20 source groups and 200 anchors, including only 10 and 37 positives, respectively. The original five-backend pilot requested 4,000 predictions; Jev Choice returned 776 valid responses out of 800 under the study's strict numerical policy, blocking its full-coverage quality comparison. On XD captions, Jev Noul achieved 75.99% average precision versus 48.47% for Qwen generated probability and 57.81% for the stronger local ordinal-likelihood expectation. The latter paired difference was 18.18 percentage points (95% source-group bootstrap interval 5.53-31.50). UCF did not show a corresponding advantage: caption ROC-AUC was 52.26% for Noul and 65.95% for ordinal likelihood. Both probability readouts had higher, hence worse, UCF Brier scores than the evaluation-prevalence reference of 0.0475. We additionally audit historical LAVAD scores at exactly matched anchors and distinguish response structure from numerical consistency. A binary-likelihood control is missing. These exploratory offline results characterize ranking, probability quality and interface failures; they establish neither a causal typed-interface benefit nor general superiority, calibration or end-to-end acceleration.