A face editor built on an Artificial Neural Network using React.js and Tensorflow.js.
Demo works best on desktop!
Built using an autoencoder, with the following architecture.
Then the model was split into the encoder and decoder. For the client app only the decoder is served. The inputs are transformed using PCA in order to extract high variance explaining components. As a result sliders have been created that change the values in the PCA components. Hence a very basic editor has been created.
For training data from Data
Rasmus Rothe and Radu Timofte and Luc Van Gool.
Deep expectation of real and apparent age from a single image without facial landmarks.
International Journal of Computer Vision (IJCV)
July 2016
Rasmus Rothe, Radu Timofte, and Luc Van Gool.
Dex: Deep expectation of apparent agefrom a single image.
InIEEE International Conference on Computer Vision Workshops(ICCVW),
December 2015.
Data required some preprocessing to unify image dimension and image quality. OpenCV was used to detect faces and extract images with faces in them.
- The neural network was trained using Keras.
- Client side serving of the network is handled by Tensorflow.js
- The app is built using React.js.