caffe & python2
1.Download the 300W dataset and run create_raw_300W_equal_bbox_train.py to obtain the annotation.
2.Prepare the dataset using generate_h5_300w_scale_V4.py ; This performs data augmentation and generates the input for the Caffe.
3.cd train_300w_68_vanilla and Run train.sh
you can download the 300W from here： https://github.com/FaceGg/DataSets
Thanks for https://github.com/cunjian/face_alignment