Abstract Drone technological advancements, driven by advances in computer vision and deep learning, have made a variety of applications in urban planning, security, and surveillance possible. Effective face detection and reliable obstacle avoidance are two crucial facets of drone operations that are covered in this study. This model is appropriate for drone platforms with limited resources since it exhibits good accuracy and real-time performance. The major idea is to implement a dual-purpose framework integrating real-time face detection and autonomous collision avoidance for UAV operations. The YOLOv8 model, fine-tuned on the WIDER FACE dataset, is employed for robust face detection, overcoming challenges such as occlusions and varying head poses, and to ensure safe navigation, a stereo camera-based obstacle avoidance system is introduced as a cost-effective alternative to LiDAR. The system utilizes Semi-Global Block Matching (SGBM) and Block Matching (BM) algorithms to generate dense disparity maps, enabling precise per-pixel depth estimation. Together, the advances improve drone-based surveillance systems by incorporating collision avoidance and real-time face detection.