Deep Alignment Network: A convolutional neural network for robust face alignment

This is a Tensorflow implementations of paper "Deep Alignment Network: A convolutional neural network for robust face alignment". You can see Original implementation here.

---

System

• No Windows !

Getting started

• Tensorflow 1.7.0
• OpenCV 3.1.0 or newer

Train Model

• Download Datasets.
• Put images & pts in SAME folder.
• Write mirror file. There is a 68 landmark mirror file. download
• Preprocess.

python preprocessing.py --input_dir=... --output_dir=... --istrain=True --repeat=10 --img_size=112 --mirror_file=./Mirror68.txt

• Train model.

python DAN_V2.py -ds 1 --data_dir=preprocess_output_dir --data_dir_test=...orNone -nlm 68 -te=15 -epe=1 -mode train
python DAN_V2.py -ds 2 --data_dir=preprocess_output_dir --data_dir_test=...orNone -nlm 68 -te=45 -epe=1 -mode train

Eval Acc

• Download Datasets for test.
• Put images & pts in SAME folder.
• Preprocess.

python preprocessing.py --input_dir=... --output_dir=... --istrain=False --img_size=112

• Eval model Acc.

python DAN_V2.py -ds 2 --data_dir=preprocess_output_dir -nlm 68 -mode eval

Results on 300W

• Speed : 4ms per Image on GTX 1080 Ti
• Err : 1.34 % on 300W common subset(bounding box diagonal normalization).

Pre-trained Model

TODO:You can download pre-trained model here. This model trained on 300W dataset.