Chen-Wei Chang

AI/ML Application Engineer
TSMC, VNAP Department

Chen-Wei Chang

I am currently working at TSMC in the VNAP department as an AI/ML Application Engineer. I received my Master's degree from the Institute of AI Innovation at National Yang Ming Chiao Tung University, advised by Prof. Yu-Lun Liu and Prof. Jiun-Long Huang, and my Bachelor's degree in Electronic Engineering from National Taipei University of Technology. My research interest focuses on Camera ISP, Generative AI, and 3D scene reconstruction.

Timeline

Mar. 2026 – Present

TSMC

AI/ML Application Engineer, VNAP

Jul. 2023 – Aug. 2025

National Yang Ming Chiao Tung University

M.S., Institute of AI Innovation

Feb. 2023 – Jun. 2023

Mores Tech

Full-Stack Engineer

Sep. 2019 – Jun. 2023

National Taipei University of Technology

B.S., Electronic Engineering

News

Mar 2026 Started working at TSMC as an AI/ML Application Engineer in the VNAP department!
Dec 2025 Honored to receive the TAAI 114 Master's Thesis Award!
Sep 2025 GCC demo now live on Hugging Face Spaces! Try it here!
Aug 2025 Served as a speaker at the AI Top Conference Paper Sharing Session organized by AICOE.
Mar 2025 Two papers GCC and SpectroMotion accepted at CVPR 2025!

Publications

GCC: Generative Color Constancy via Diffusing a Color Checker

Chen-Wei Chang, Cheng-De Fan, Chia-Che Chang, Yi-Chen Lo, Yu-Chee Tseng, Jiun-Long Huang, Yu-Lun Liu

CVPR 2025

Proposed GCC (Generative Color Constancy), a novel approach that leverages pretrained diffusion models (Stable Diffusion) to estimate scene illumination by generating color checkers into images, enabling auto white balance across different camera sensors without requiring camera-specific training.

SpectroMotion: Dynamic 3D Reconstruction of Specular Scenes

Cheng-De Fan, Chen-Wei Chang, Yi-Ruei Liu, Jie-Ying Lee, Jiun-Long Huang, Yu-Chee Tseng, Yu-Lun Liu

CVPR 2025

SpectroMotion combines 3D Gaussian Splatting with physically-based rendering (PBR) and deformation fields to reconstruct dynamic specular scenes, and is the only existing 3DGS method capable of synthesizing photorealistic real-world dynamic specular scenes.

Projects

DS3DGS

DS3DGS: Dynamic Specular 3D Gaussian Splatting

Our proposed approach, DS3DGS, expands 3DGS by adding a deformation field with specular surface rendering, significantly improving the quality of dynamic specular scene reconstruction.

VAE Sentiment Analysis

VAE based Sentiment Analysis

Employing a Variational Autoencoder (VAE) model to perform sentiment analysis by mapping sentence embeddings, derived from a pre-trained BERT model, into a latent space composed of Arousal and Valence dimensions.

Game Platform

RESTful Web Game Platform

Developed an interactive web gaming platform featuring user authentication, two classic games (Whac-A-Mole and Snake), and a real-time leaderboard system implemented with RESTful APIs.

Golf System

Golf Ball Trajectory Estimation System

This project focuses on developing a golf trajectory estimation system that uses high-speed cameras and image processing algorithms to calculate and simulate golf ball trajectories.