Enhanced Abnormal Activity Detection: Utilizing YOLOv8 and Deep SORT with TSAI and LSTM Classifiers

Video analysis has undergone a radical transformation thanks to the quick development of deep learning and computer vision techniques, especially in the area of anomalous activity detection. Robust abnormal activity recognition algorithms are a prerequisite for strong security and safety protocols, which has led to the investigation of novel approaches. This work examined the combination of two state-of-the-art classifiers—Long Short-Term Memory, or LSTM, and Time Series AI, or TSAI—with cutting-edge techniques for object tracking and recognition, such as YOLOv8 (You Only Look Once, version 8) and Deep SORT (Simple Online and real-time tracking). The suggested method attempted to precisely identify anomalous activities in video feeds by utilizing the skills of LSTM and TSAI in capturing temporal dependencies and analyzing sequential data, along with the real-time object tracking abilities of Deep SORT and accurate object detection of YOLOv8. The performance of the integrated system was evaluated utilizing precision, recall, and F1-score metrics during an extensive examination that covered a variety of datasets and situations, resulting in an astounding 97.22% accuracy. With the ultimate goal of improving the system’s practical usefulness in real-world settings across industries like retail, disaster management, healthcare, and building security, the study also investigated the effects of various configurations and parameters on the system’s efficacy. By improving video analysis techniques, this research helps enterprises improve security protocols and successfully maintain public safety.