My research interests are machine learning, computer vision, biomedical image analysis, healthcare, interpretability of machine learning, and visual analytics.
I am advised by Anthony Gitter and Yu Hen Hu. My projects include fluorescent cell image classification and drug discovery with cell painting data. Previously, I worked on video object tracking and facial reenactment on mobile devices.
Cell Image Classification
Advisor: Anthony Gitter, Dec. 2017 - Present
- Classify T-cell and breast cancer cell types using fluorescent images
- Compare and analyze various classifiers: logistic regression, fully connected
neural network, convolutional neural network and transfer learning
- Interpret feature representations from different learning layers in the transfer learning model
Cell Painting and Drug Discovery
Advisor: Anthony Gitter, July 2018 - Present
- Analyze about 1 million 5-channel cell-painting images of bone tumor cells
- Explore latent space between the image space and chemical molecule space
- Study how to remove batch effects of microscopy images at scale
Video Object Tracking
Advisor: Yu Hen Hu, Feb. 2017 - Dec. 2017
- Study how to track car driver’s head position and orientation from low-quality traffic video
- Develop semi-automatic video annotation software with Viola-Jones frontal face detector for training object tracking algorithms
Advisor: Yu Hen Hu, June 2017 - Dec. 2017
- Implement real-time face tracking algorithms on iOS devices
- Train a facial reenactment model using GANs and port it to iOS devices
Using transfer learning to classify breast cancer cells with fluorescence imaging
Zijie (Jay) Wang, Tiffany Heaster, Quan Yin, Alex Walsh, Melissa Skala, Anthony Gitter
In UW–Madison Undergraduate Symposium '18.
Selected Course Projects
Iteractive 2D, 2.5D and 3D computational graphics with shaders and textures, created by using HTML canvas and webGL.
Sentiment analysis: predicting Yelp ratings based on text comments of Madison restaurants. This project won the in-class Kaggle competition.
Flexible and robust Mixed Integer Quadratic Programming model written in Julia to solve a real-life group assignment problem. It won the best project.
Convenient macOS command line tool for uploading screen-shots from the clipboard to Imgur. This app is featured on the official Imgur website.
MacOS GUI used to auto-annotate facial landmarks from a video. These landmarks can further be used to train GANs.
Interactive geographic visualization for studying where UW–Madison Chinese undergraduates are from.