Education

Ph.D. in University of Maryland, College Park
Aug 2021 - Now
Ph.D. in Computer Science
Advised by Prof. Jia-Bin Huang.

B.S. in National Cheng Kung University
2008 - 2013
Electrical Engineering (EE), Industrail and Information Management (IIM), double major
EE advised by Prof. Soon-Jyh Chang
IIM advised by Prof. Chiang Kao

Publication

Work Experience

  • Neural radiance fields related topics.
  • Conducted research on multimodal late fusion for object detection.
  • Conduct research on 3D object detection, connecting 2D/3D features for better recognition and box localization.
  • Achieved 3% mean average precision (mAP) gain for indoor scenes.
Italian Trulli
  • Utilized fine-grained structure of face in feature space for accurate head pose estimation, resulting a fast and compact CNN model.
  • Disentangled the information of image style and person classification features for person re-identification, and verified the disentanglement with cycle consistency of Generative Adversarial Network (GAN) using Pytorch.
  • Few shot learning for image classification
  • Established algorithm to enhance image/video contrast that works with low computational cost and high flexibility for cell phone chips.
  • Developed scene recognition algorithm to assist with camera auto-exposure and auto-white-balance functions, raising the correctness of color assignment.
  • Implemented universal auto-white-balance calibration approach that eliminated difference between different modules, saving time of module calibration.
  • Designed and tested camera auto-exposure algorithm for HDR images, enhancing overall image quality which met the requirements from customers.

Course Project

MSCV Capstone Project

Object Detection in Infrared Images

In this project, we focus on detection with RGB and infrared signals in the driving scenes.
[Website]
Italian Trulli

Course: 16720 Computer Vision - B

Rectangling Panoramic Images via Warping

We implemented the paper: Rectangling Panoramic Images via Warping, SIGGRAPH 2013, as our final project of course 16720B. You can find the code below.
[Code] [Slides]