Cigarette detection in images and videos has emerged as a challenging problem in computer vision with diverse applications in public health, regulatory compliance, and behavioral monitoring. This chapter presents a novel approach for offline video-based cigarette detection, addressing the inherent complexities associated with this task. In recent years, researchers have adapted advanced object detection techniques, to detect cigarettes in various contexts. However, the scarcity of annotated datasets specifically designed for cigarette detection remains a significant hurdle. Cigarettes are often small and inconspicuous objects, and they are frequently found in dynamic and cluttered scenes, making their detection a formidable challenge. To tackle these challenges, we proposed an approach consisting of a multistep workflow that includes human detection, body part detection, and cigarette classification on hand and mouth crops. We have rigorously assessed our proposed cigarette detection framework within the context of Movies and TV Shows, providing an evaluation of its performance on a real-world problem.