In our project, we used SIFT algorithm to extract the image features and RANSAC algorithm for feature matching. We mainly focused on finding the best performance of SIFT + RANSAC face recognition system under different conditions. Each image is a metadata and has eight different features (date, gender, race, year, weather, glasses, expression, yaw). For a normal face recognition scenario in a public dataset, all of these features are useful and distinguishable for recognizing face. It becomes, however, way more challenging to recognize (identical) twin faces only using race or gender information.
For more information please see our paper