Does your face reveal how long will you live?

Recent advances in deep neural networks, on both the software and the hardware front, have increased their accessibility and applicability. A key area of growth is the use of neural networks in prediction of health related endpoints such as mortality or morbidity.

In this talk, I will describe my work on extracting age related health information from facial imagery. In particular, I will show how we leverage transfer learning to train models for the data-rich task of age prediction, that we later fine-tune to predict perceived age and mortality risk. I will also discuss importance of preprocessing, augmentation, and data privacy.