Audio-Visual Dataset and Method for Anomaly Detection in Traffic Videos
This work addresses traffic safety by enhancing anomaly detection with multimodal data, though it is incremental as it builds on existing anomaly detection methods.
The authors tackled traffic anomaly detection by introducing MAVAD, the first audio-visual dataset from real-world scenes with diverse conditions, and proposed AVACA, a novel method combining visual and audio features via cross-attention, which improved performance by up to 5.2% with audio addition.
We introduce the first audio-visual dataset for traffic anomaly detection taken from real-world scenes, called MAVAD, with a diverse range of weather and illumination conditions. In addition, we propose a novel method named AVACA that combines visual and audio features extracted from video sequences by means of cross-attention to detect anomalies. We demonstrate that the addition of audio improves the performance of AVACA by up to 5.2%. We also evaluate the impact of image anonymization, showing only a minor decrease in performance averaging at 1.7%.