This document discusses a project that uses deep metric learning techniques for human face detection and identification in images and videos. Deep metric learning outputs a real-valued vector rather than a single classification. It uses libraries like OpenCV, Dlib, scikit-learn and Keras to build neural networks for facial recognition. The goals are to develop a system that can identify faces even from low quality images with variations in illumination, expression, angle and occlusions. Existing face recognition has challenges in these conditions, so the aim is to improve accuracy rates for normal and non-ideal images through deep metric learning approaches.