Fine-tuned FaceNet model
We fine-tuned the open-source FaceNet model (version 20170512-110547), which is trained on the MS-Celeb-1M dataset, with the VGGFace2 dataset via a combination of softmax and center loss.
The detail implementation is provided in log dir.
We also provide modified train_softmax.py that allows you to perform fine-tuning with FaceNet model