It is a neural network model that detects 15 facial points. * The Dataset : Kaggle Dataset: https://lnkd.in/fp9nhXg • The Model : Created a basic NN using Keras sequential containing 1 cov2d and 1 max pool and 2 hidden layer (128 and 64) and output layer (30) • Training accuracy : 74-75 percent • The Detector and Predictor : I used opencv Haar cascade to detect faces. Loaded the above created model(h5 file) to create a instance of model. Faces based on the detector are feeded to the model.predict() function to get the prediction and (15)key points.
Don't forget to tag @shubham-chhimpa in your comment, otherwise they may not be notified.