rahul-pande / faces-mtl

Multi Task Learning example with Keras

faces-mtl

Multi Task Learning example with Keras

Multitask learning is powerful when the tasks could benefit from having shared low-level features. For example predicting the age and gender are different tasks, one being regression and the other being classification. However, solving these tasks simultaneously makes the network learn better low-level representations of the face than solving these taks separately.

Here we use cropped face image data set from https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/. Each image is preprocessed to be of same size.

We have used Functional API of Keras to implement multi-output neural network. we have a cross entropy loss for the gender output and RMSE for age output.

Below is the model architecture.

All code and documentation is present in the jupyter notebook. Contact [e-mail] for questions.