Eric Z. Chen

I will be pursuing my Master of Science in Computer Science at the University of Illinois Urbana-Champaign (UIUC) starting in Fall 2024, and my interests are in computer vision and multimodal learning and generation.

Previously, I graduated with a B.S. in Statistics & Computer Science from UIUC in December 2023.

Email  /  LinkedIn  /  GitHub  /  Google Scholar

UIUC Yuxiong Lab, advised by Prof. Yuxiong Wang

I have been working alongside Jiahua Dong and Vedaant Jain on expanding and improving the results from the previous YouTubePD paper. My work has been to simulate Parkinson's symptoms in the faces of healthy individuals at the video level using diffusion-based video editing methods, and at the image level using ControlNet inpainting to generate augmented PD-positive training images. Additionally, I have been tuning our model to improve metrics.

Google, with Junyi Zhang and advised by Prof. Ming-Hsuan Yang

I worked with Junyi Zhang and the research team on improving the results from their previous SD-DINO paper, by utilizing the new Stable Diffusion XL model to improve previous semantic correspondence performance. My work was to extract features from SDXL to learn about its geometric understanding of images, help integrate SDXL into other baselines and run experiments to evaluate their metrics, and assist in creating and processing a new semantic correspondence benchmark.

Publication:
Junyi Zhang, Charles Herrmann, Junhwa Hur, Eric Chen, Varun Jampani, Deqing Sun, Ming-Hsuan Yang. Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence. Conference on Computer Vision and Pattern Recognition (CVPR) 2024. Project Page, Arxiv

UIUC Forward Data Lab, advised by Prof. Kevin C.C. Chang

From Summer 2022 to Summer 2023, I worked on the Living Encyclopedia project, a web application that automatically generates concise and accurate wiki pages containing general and academic information for university departments. This project included BERT fine-tuning for classification, active learning, summarization, information retrieval, feature engineering, and text data augmentation techniques.

In the Fall 2023 semester, I explored prompt engineering methods to create suitable prompts for citation-aware summarization and intelligent content extraction from web pages, as additions to the Living Encyclopedia project. I also conducted experiments to benchmark Llama 2's performance in various tasks, such as custom extraction, summarization, and classification.

PyTorch implementation of the Neural Style Transfer paper by Gatys et al., using a pre-trained VGG-19 convolutional neural network and calculating the style and content losses from scratch.

Streamlit Demo

• Intramural Basketball, Intramural Soccer, Badminton Club
• ACM@UIUC SIGAIDA, AAA

• Weightlifting, Calisthenics
• Sports, watching sports
• Exploring Nature