Facial Detection and Recognition in Drone Imagery Using FaceNet

Face recognition in drone-captured images has emerged as a promising technology with diverse applications, including surveillance, security, search, and rescue operations. Recognizing faces from heights has limitations because the images obtained by drones may have varying angles, distances, lighting conditions, occlusions, and in some cases, only small portion of the face might be visible. It is crucial to consider these limitations and develop suitable techniques for reliable performance. In this paper, an efficient approach for recognizing the faces in the drone images even in the challenging conditions is proposed. For face detection, multi-task cascaded convolutional neural network (MTCNN) is employed. To extract the most important facial features, FaceNet, a deep learning model that transforms facial images into a 128-dimensional feature vector representation, is used. K-Nearest Neighbor (KNN) is used to identify a face by comparing its features extracted using FaceNet with the features of known faces in a training dataset. Experimental results demonstrate the effectiveness of our approach in detecting and classifying faces in drone-captured images. The system achieves 96.02% accuracy and robust performance.