Facial Recognition for Humanitarian Efforts: A Deep Learning based Solution for Missing Person Identification

Our work offers a novel method for identifying missing persons that makes use of recent developments in face recognition technology. Conventional techniques depend on converting images into embeddings and then figuring out how far apart picture embeddings are from one another by calculating distances between them, making it computationally demanding, particularly while dealing with huge databases. Our process involves first extracting faces from images using OpenCV’s face extractor module, and then using VGGFace to turn those faces into embeddings. However, we employ Principal Component Analysis (PCA) to further reduce their dimensionality to 128-dimensional vectors rather than directly comparing embeddings. On the basis of these reduced embeddings, a Dense Neural Network is then trained. With this method, the traditional O(n) time for classification is greatly reduced to O(1). Our tests show that our methodology outperforms numerous other methods and reaches an excellent accuracy of 98.75%. This simplified method gives encouraging outcomes in situations involving missing persons identification in addition to increasing efficiency.