DaoAI needed specialized tools to determine 3D positions of objects on conveyer belts using stereo cameras, and to estimate their 6D poses, so that a manufacturing robot can be guided to the correct 3D space with the ”hands” in the correct orientation to grab the object.
I re-implemented well-known stereo matching and object pose estimation methods. I collected datasets, fine-tuned the machine learning models, and re-created state-of-the-art results from the respective publications. The tools used were Python, OpenCV, Open3D, and PyTorch.
These essential tools could then be offered to clients as needed, helping speed up and automate more of the manufacturing processes.
Teaching Assistant
Simon Fraser University
TA for CMPT115: Exploring Computer Science. (Summer 2022 & Summer 2021)
TA for CMPT125: Introduction to Computing Science and Programming II. (Spring 2022 & Summer 2020)
TA for CMPT120: Introduction to Computing Science and Programming I. (Summer 2020)
Undergraduate Research Fellow
University of Hong Kong
Conducted research in Computer Vision as part of a prestigious programme (Undergraduate Research Fellowship Programme) at HKU.
Research topic: “Matting Colored Transparent Objects from a Single Image Using Deep Learning.” (Paper on arxiv and on my homepage)
Won the best Final Year Project award.
Education
MSc Computing Science
Simon Fraser University
GPA: 3.87/4.33
Worked under the supervision of Dr. Ping Tan in the field of Computer Vision and Artificial Intelligence. Focus on 3D vision.
Thesis title: “SAgA-NeRF: Subject-Agnostic and Animatable Neural Radiance Fields for Human Avatar”