Amy Zhao

About Me

I am a PhD student at MIT working on computer vision and machine learning. I am advised by Professors John V. Guttag and Frédo Durand. I am interested in modeling the transformations that we observe in realistic images, including 3D rotations of objects, complex lighting effects and even artistic effects. These models have a variety of applications including data augmentation and image synthesis.

Prior to graduate school, I worked for 3 years as a Software Engineer at Microsoft and Caradigm. I did my undergraduate studies at the University of Toronto in Engineering Science and Biomedical Engineering.

Recent News

Our work on data augmentation for medical image segmentation was accepted to CVPR 2019 as an oral presentation!

Check out some recent publications about VoxelMorph!

Faster analysis of medical images | MIT News

Algorithm speeds up medical diagnosis | Innovators Magazine

Machine learning could enable medical image registration during operations | Radiology Business

Recent Publications

Data augmentation using learned transforms for one-shot medical image segmentation

A. Zhao, G. Balakrishnan, F. Durand, J. Guttag, A. V. Dalca.

CVPR 2019 (oral and poster)

Project page Paper

We learn independent models of spatial and appearance transforms, and use them to synthesize new training examples for a supervised segmentation network.

VoxelMorph: A Learning Framework for Deformable Medical Image Registration

G. Balakrishnan, A. Zhao, M. R. Sabuncu, J. Guttag, A. V. Dalca.

IEEE Transactions on Medical Imaging 2019

Project page Paper

An Unsupervised Learning Model for Deformable Medical Image Registration

G. Balakrishnan, A. Zhao, M. R. Sabuncu, J. Guttag, A. V. Dalca.

CVPR 2018 (poster)

Project page Paper

We train an unsupervised model called Voxelmorph that registers medical images as accurately as the state-of-the-art, but orders of magnitude more quickly.

Synthesizing Images of Humans In Unseen Poses

G. Balakrishnan, A. Zhao, A. V. Dalca, F. Durand, J. Guttag.

CVPR 2018 (oral and poster)

Project page Paper

We use a modular neural network architecture to simplify the image synthesis problem, producing high-quality images that preserve the input person's appearance.

A Video-Based Method for Automatically Rating Ataxia

R. Jaroensri*, A. Zhao*, G. Balakrishnan, D. Lo, J. D. Schmahmann, F. Durand, J. Guttag.

Machine Learning for Healthcare 2017

Paper

We use pose estimation and feature engineering to produce ratings of neurological movement disorder severity from videos of motor exams.