"Facial landmark detection is the task of detecting key landmarks on the face and tracking them (being robust to rigid and non-rigid facial deformations due to head movements and facial expressions)"
- Image --> CNN --> output co-ordinates of the landmarks
- Explained by Andrew Ng
- further see
facial-keypoints.ipynb
for seeing the training process and network architechture
- Since the data set only contain image of size 96,96,1
- Hence haar-cascade is used to first isolate the face and that is passed into model
image.py
runs the network on any imagevedio.py
runs the network on vedio from web cam
- Better dataset ( or clean this one)
- Data contains only some images with 15 landmarks , rest contain only 4
- Better model
- https://www.kaggle.com/mananmadan/facial-keypoint-detection