Vending machines are part of everyday life in public places such as transit stations, hospitals, and college campuses. Even today, most of them depend on buttons or touch screens. This leads to frequent surface contact, regular wear, and usability issues for some users. It also becomes a concern in crowded or hygiene-sensitive environments. Although hand gesture recognition has been widely studied, its use in practical public systems remains limited. This paper presents a touch-free vending interaction system that allows users to browse and select products using simple hand gestures. A camera situated near the vending area records live footage, which is then processed by a YOLO-based vision model. The model detects the hand and recognizes the gesture in a single step, allowing the system to reply rapidly without requiring extensive preprocessing. Instead of treating gestures as separate inputs, the system analyzes them based on the current state of involvement. This helps to prevent unintended actions and makes the whole flow more predictable. Only a small set of gestures is used so that first-time users can interact comfortably without prior instruction. Training begins with publicly available gesture datasets and is later adjusted using a small dataset collected to reflect real vending usage, including changes in lighting and hand position. Unreliable detections are deleted while in operation to maintain consistent performance. This study focuses on both recognition accuracy and practical use. The proposed approach effectively demonstrates the implementation of gesture-based control in public vending machines by integrating real-time gesture detection into a unified interaction framework, ensuring simplicity and usability.
S et al. (Thu,) studied this question.