this repositorie shows a regression for detect landmark face using a Convolution Neural Network. Helen dataset is used for train a CNN. HoG model is used for face detection. Model is deployed on PC with Python and on Android app.
Step 1: this notebook plots some images and landmark face.
Step 2: this notebook scales the images (and landmarks) and converts to gray scale.
Step 3: this notebook normalizes the images, trains a CNN using Keras, and saves a TensorFlow model.
Step 4: this notebook tests regression model. Input to model comes from image file.
Step 5: this Python script tests the model on PC with Python. Input to model comes from PC webcam.
Step 6: this notebook transforms H5 TensorFow model to TensorFlow Lite, this model is deployed on Android app.
