Open source implementation of the renowned publication titled "DeepFace: Closing the Gap to Human-Level Performance in Face Verification" by Yaniv Taigman, Ming Yang, Marc'Aurelio Ranzato, Lior Wolf published at Conference on Computer Vision and Pattern Recognition (CVPR) 2014.
Implementation of this paper have been done using Keras (tf.keras). This project and necessary research was supported by the TensorFlow Research Cloud (TFRC). GCP resources have been used to train a million-scale machine learning model using Cloud TPUs.
Make your account on http://zeus.robots.ox.ac.uk/vgg_face2/login/ once you login goto http://zeus.robots.ox.ac.uk/vgg_face2/ and download all the files mentioned there
- Train Data_v1. 36G. MD5: 88813c6b15de58afc8fa75ea83361d7f.
- Test Data_v1. 1.9G. MD5: bb7a323824d1004e14e00c23974facd3.
- Train_Images_v1. The training image list, e.g., 'n000002/0001_01.jpg'.
- Test_Images_v1. The test image list.
If are using a server a quick way to get this data is hijack it using cookie, for example:
curl --header "cookie: " http://www.robots.ox.ac.uk/~vgg/data/vgg_face2/meta/test_list.txt --output /home/Documents/ajinkya/output/test_images_v1.txt
After you have all this data run python3 generate_train_labels.py make sure you enter the path of 'train' folder that you downloaded in previous step. This code will generate VGGFace2-class_labels_train.txt file which you need in next steps
Goto prepare_tfrecords.py file from tfrecords-faster directory and run it make sure of the paths and generate tfrecords for both train and test