Tensorflow implementation of SSPP-DAN: Deep Domain Adaptation Network for Face Recognition with Single Sample Per Person
First, download the dataset or the pickle files that we have already created from our repository. After all pickle files are download, move them (i.e., src_train.pkl, target_test.pkl, target_train.pkl) into the SSPP-DAN/data folder.
Then run get_vggface.sh in the SSPP-DAN/pretrained folder to use the pre-trained VGG-Face model.
To train a model with downloaded dataset:
$ python train_model.py --learning_rate=1e-5 --batch_size=50 --save_step=100
To test with an existing model:
$ python test_model.py --summaries_dir 'expr/F3D_30_60_FC6_FC6' --test_batch_size=50
Facial feature space (left) and its embedding space after applying DA (right). The subscript “s” and “t” in the legend refer to the source and target domains, respectively.
Sungeun Hong / @[csehong][wp] [wp]: sites.google.com/site/csehong