Human Gesture Dataset for UAV Control

This dataset contains labeled images for human gesture recognition applied to Unmanned Aerial Vehicle (UAV) control using computer vision and deep learning techniques. The dataset is organized in a single directory named dataset, where each subfolder represents a distinct class corresponding to a control command. The classes included are: forward backward left right up down stopped w_pee x Each subfolder contains images of a human operator performing the respective gesture. The dataset was collected in real-world environment, including variations in lighting conditions, backgrounds, distances, and execution styles, aiming to improve robustness in practical UAV applications. The dataset is designed for image classification tasks, particularly using deep learning models such as YOLOv8 (classification mode), and supports research in: Human–Drone Interaction (HDI) Gesture-based control systems Computer Vision for UAV navigation Real-time embedded AI systems Special classes such as w_pee (contextual/no-PPE condition) and x (undefined or irrelevant actions) are included to improve model generalization and safety by allowing the system to correctly identify non-command situations. This dataset was used to develop and validate a real-time gesture recognition system integrated with a UAV, enabling intuitive and safe human–robot interaction.