Self-restrained-Triplet-Loss
Self-restrained-Triplet-Loss copied to clipboard
Official repository for Self-restrained Triplet Loss for Accurate Masked Face Recognition
I am trying to reproduce the results using given EUM weights, Resnet100 pre-trained weights downloaded from https://github.com/deepinsight/insightface/tree/master/model_zoo and LFW mask type and colors given in LFW-log.txt. But my results are...
hello there! thank you so much for sharing your amazing work. would you please assist me how can i prepare my own custom dataset and what should the the directory...
Are the pretrained models used for feature extraction trained under masked face dataset or normal ones?
lfw_eval
Hello, i tried to run evaluation on lfw. When i run align_lfw.py, it shows im missing 'shape_predictor_68_face_landmarks.dat'. Where can i get this file? thanks
Please share python environment required packages in requirements.txt.
I am unable to run the faceMasked.py file. Getting "1.jpg is not in the list" despite tried to re-install the logs of the IJB-C from the repo to the local...
Where can I download the model defined in face_model.py to preprocess the embedding of the input images? It seems that only the weights of fine-tuned cnn models are available for...
Hello, Can you provide 'ijbc_name_box_score_5pts.txt' used by evaluation/ijbc/FaceMasked.py plz. My version of IJB-C dataset doesn't have the exactly same file.
In the section 4.5 of the paper: > All models are trained using an SGD optimizer with an initial learning rate of 1e−1 and batch size of > 512. The...
https://github.com/fdbtrs/Self-restrained-Triplet-Loss/blob/adb9421fb8171b530d1351a59763c3be6f0fd346/main.py#L141 It should be `cnn_optimizer.zero_grad() `