Learning Traffic Anomalies from Generative Models on Real-Time Observations

TL;DR

This paper presents a novel approach using a Spatiotemporal GAN with Graph Neural Networks and LSTMs to detect traffic anomalies from real-time camera data, achieving high accuracy and low false positives.

Contribution

The study introduces a new deep learning framework combining GNNs and LSTMs within a GAN for real-time traffic anomaly detection from visual data.

Findings

High precision detection of traffic anomalies

Effective identification of signal interruptions and weather effects

Low false positive rates in anomaly detection

Abstract

Accurate detection of traffic anomalies is crucial for effective urban traffic management and congestion mitigation. We use the Spatiotemporal Generative Adversarial Network (STGAN) framework combining Graph Neural Networks and Long Short-Term Memory networks to capture complex spatial and temporal dependencies in traffic data. We apply STGAN to real-time, minute-by-minute observations from 42 traffic cameras across Gothenburg, Sweden, collected over several months in 2020. The images are processed to compute a flow metric representing vehicle density, which serves as input for the model. Training is conducted on data from April to November 2020, and validation is performed on a separate dataset from November 14 to 23, 2020. Our results demonstrate that the model effectively detects traffic anomalies with high precision and low false positive rates. The detected anomalies include camera signal interruptions, visual artifacts, and extreme weather conditions affecting traffic flow.