Personality-disentangled-FER

This is a personality-disentangled FER(Facial Expression Recognition) project launched together by Yi Huo and Amir Shirian. The first code 'extractFras-20230202.py' is to process dataset(https://chalearnlap.cvc.uab.cat/dataset/24/description/#). It extracts multiple image frames for each identity video. The second code 'geneFacepairCSV-face pair generation & save pairs to csv file-20230205.py' is to generate face image pair combinations for evaluating the the performance of various models on face verification. One point to emphasis is that its delete repetitive identity that distributed in different higher level dirs than the first code used 1-order dirs. This is, the first code delete repetitive identity videos, while this second code delete repetitive extracted identity framesfolders further. The third code 'face verify-default(VGGFace)-20230207.py' run on a list of fair image pairs to verify whether each image pair is from a same person. After testing all the face pairs, it will calculate measurement results including TP, TN, FP, FN, Recall, Precision and Accuracy. This code is used to test different face recognition models on their performance about face verification.