Shoplifting Detection in Video Surveillance Using SlowFast
Author : Gagana TV and Asha S
Abstract :
Shoplifting is a major problem for retail stores, resulting in financial losses and operational inefficiencies. Traditional surveillance systems, while useful for monitoring activities, often require constant human supervision, making them prone to oversight and delays in response. To address this issue, a Shoplifting Detection System has been developed to automatically analyze surveillance videos and identify suspicious behavior. The system processes the uploaded video by extracting frames at regular intervals and analyzing them to detect the instances of shoplifting. When a shoplifting activity is detected, the system triggers an alarm, alerting security team to take immediate action. This approach enhances the security by reducing human error and ensuring quicker responses to such threats. The system follows a structured workflow where the video is first uploaded for analysis, followed by frame extraction and processing to identify unusual activities. When shoplifting is detected, the system suddenly activates the alarm to notify the security team. By automating the detection process, this system significantly reduces the need of manual monitoring, allowing security team to focus on other critical tasks. This not only improves overall store security but also helps prevent potential losses, making the monitoring process more efficient and reliable.
Keywords :
Accuracy, SlowFast, DCSASS Dataset.
