RoadCrime-CCTV: AI-Enabled Roadside Crime Detection Dataset

This dataset contains 219+ CCTV video samples and extracted high-resolution image frames capturing real-world roadside and intersection criminal behaviors, specifically theft, robbery, and snatching in Bangladesh. Designed to support automated, proactive threat detection using Computer Vision and Machine Learning, the dataset covers temporal sequences categorized into three distinct phases: Pre-event (suspicious behavior before the crime), Event (crime in progress), and Post-event (escape or crowd reaction). The footage reflects natural urban and highway environments, capturing variations in traffic densities, pedestrian movements, and lighting or weather conditions. It aims to bridge the gap in existing resources by providing crucial pre- and post-violation context for behavioral analysis and intelligent surveillance integration. Subject Areas: Computer Science, Road Safety, Criminology. Keywords: Road Crime Detection, CCTV Dataset, Computer Vision, Machine Learning, Smart City Surveillance, Behavior Analysis. Data Type: Image, Video.