search

yinguobing / cnn-facial-landmark

Training code for facial landmark detection based on deep convolutional neural network.
MIT License
623 stars 183 forks source link

Use z coordinate to get 3D prediction #68

Open abdou31 opened 4 years ago

abdou31 commented 4 years ago

Hello Yin, I would like to use z coordinate and add it to the pts file. Is it possible to train dataset (images + .json files ) with x, y and z coordinates with the CNN that you created? For e.g, if landmarks that I want to use are 40 landmarks, units in logits layer should equal to 120.

yinguobing commented 4 years ago

I think it's possible, and in that case the number of node in the output layer is 120 as you said.

abdou31 commented 4 years ago

So , pts file should be in this form :

n_points {
9.14523 5.2145 1.365
1.2541 1.254 1.879
}

But what should I change in pts_tools.py and extract_points_from_ibug.py?