About Me

My Name is Hengjia Zhang (张恒嘉). Currently I’m a master student majoring in Computer Science and Engineering at the Univeristy of Michigan and planning to graduate in May, 2020. I got dual degrees in Computer Science at the University of Michigan and in Electrical and Computer Engineering at Shanghai Jiao Tong University. I’m a Michigan Fan, Golden State Warrior Fan, LEGO Fan and LPL Fan!

My main interest is machine learning and deep learning. I was a member of Michigan Database Research Group and of GEMS LAB.

Last semester, when I was taking deep learning course by Honglak Lee, our final project was the convertion/translation between real person images and anime images based on CycleGAN. The main modifcation was that we changed the architecture of generator and discriminator: for discriminator, we added dialtion convolutional layers to better extract global image features and add skip connections on generator so that it could preserve more image details.

When I was an undergraduate, one of my research projects was a JAVA code autocompletion based on machine learning algorithms and visualization through editor ATOM. Its final goal was trying to develop a system which could write the code itself!

Also, my team and I developed a gesture recognition system under complicated background with an accuracy of 100% and response time of 60 milliseconds per frame for Panasonic. We used the gesture as game input to prove that it’s robust and fast.

Right now I’m a deep learning intern at the Mathworks and working on pre-trained Keras Model transformation in MATLAB. Before that I was a Graduate Student Instructor at UM for EECS 486 Web Search and Information Retrieval, where I led discussion session, held office hour and created exams for the students.