Master of Science
The widespread adoption of city surveillance systems has led to an increase in the use of surveillance videos for maintaining public safety and security. This thesis tackles the problem of detecting anomalous events in surveillance videos. The goal is to automatically identify abnormal events by learning from both normal and abnormal videos. Most of previous works consider any deviation from learned normal patterns as an anomaly, which may not always be valid since the same activity could be normal or abnormal under different circumstances. To address this issue, the thesis utilizes the Two-Stream Inflated 3D (I3D) Convolutional Networks to extract spatial and temporal video features and demonstrates how it outperforms the 3D Convolutional Network (C3D) used in prior work as feature extractor. To avoid annotating abnormal activities in training videos, a weakly supervised anomaly detection model is implemented based on the Multiple Instance Learning (MIL) framework. The model considers normal and abnormal videos as bags and video clips as instances, learns a ranking model to predict high anomaly scores for video clips containing anomalies. The thesis further shows that the choice of features input, such as concatenating RGB and flow features, and careful choice of optimization settings, such as optimizer, can significantly improve the performance of the anomaly detection model on some evaluation metrics.
Summary for Lay Audience
Anomaly detection in computer vision is the task of recognizing rare or abnormal events or behaviours in videos, such as the presence of unexpected objects, changes in expected motion patterns, or deviations from the norm. It has many applications, including medical imaging, traffic monitoring, and surveillance. Video anomaly detection is essential in various applications, including medical imaging, traffic monitoring, and surveillance. Anomaly detection in surveillance videos is a vital tool for identifying potential security threats and alerting security personnel to take action. However, traditional video analysis methods rely on human monitoring, which can be prone to errors and time-consuming. Therefore, developing an automatic video anomaly detection system is crucial to reduce the need for human resources and improve the accuracy of detection. The development of a video anomaly detection system involves feature extraction from the video data, where appearance-based and motion-based features are identified and selected to differentiate normal behaviour from abnormal behaviour. Machine learning models are then trained using these features to detect anomalous behaviour in the video data. Recent advancements in deep learning have led to the emergence of new methods for anomaly detection in surveillance videos, potentially achieving superior performance compared to traditional machine learning systems. In this thesis, we proposed an anomaly detection system based on both appearance-based and motion-based features to detect anomalies happened in surveillance videos. Our model demonstrates promising results in detecting abnormal events in videos.
Soltani Nejad, Sareh, "Weakly-Supervised Anomaly Detection in Surveillance Videos Based on Two-Stream I3D Convolution Network" (2023). Electronic Thesis and Dissertation Repository. 9608.