Personal Projects

Maker

• Air Hockey Robot : uses a webcam to stream the video, process it on the fly, and feed the decision output to the robot.

• Stereo 3D Scanner​ : uses Structured-Light Stereo Imaging technique.

• 3D Printer : a DIY 3D printer based on Prusa i3 model

Microsoft

Senior Software Engineer

• Ground truth system: I currently work on various ground truth systems to validate the accuracy of the 3D reconstruction & SLAM system, the stability of the hologram, the performance of online calibration, and various metrics that are related to the AR device.

Magic Leap

Lead Software Engineer

Develop calibration process and software, propose and design methods to generate the ground truth for automatic robotic validation. Specifically the validation for the eye-tracking system, head-pose, and Magic Leap controller.

• Camera Calibration: monocular/stereo calibration to get camera intrinsics/extrinsics. They are used to generate the ground truth for validation or used to reconstruct the target in 3D.

• Hand-Eye Calibration: determine the transformation between a robot end-effector and a camera, for head-pose accuracy validation.

• Computer Vision: miscellaneous computer vision tasks, including feature extraction and object tracking.

Garmin

Senior Software Engineer

Develop the 3D engine for the next generation GPS software. Mainly focus on the Camera System, Auto Strategy, Animation System, and various mathematical intensive and 3D geometry related tasks. Some of my developed features:

• Geocentric Coordinate: make the underlying surface a true sphere instead of a plane.

• 1st-Person-View: where camera is located and pivoted at the car position, such that the camera rotates and tilts like realistic situation. This adds variety to the default 3rd-person view.

• Non-Center Camera Scheme: implemented Pinch-to-Zoom feature, where the user can focus on any position on the map, and zoom or rotate around the focus point. As opposed to the default centered mode, where the user can only rotate and zoom around the map center.

• Control System: prototyped a Control System for the camera actions. The goal of this prototype is for the camera actions (rotation, tilt, and zoom) to be controlled by some kind of system, such that the transition is not just linear, but more of the real world situation.

Structure Studios

C++ Software Engineer

Develop and maintain the 3D pool and landscaping design software, particularly in solving 2D & 3D geometry related problems and developing 3D procedural mesh engine ( by C/C++, MFC, WPF, and C# ).

• Developed several released new features for the 3D Geometry engine within 1.5 year, including:

  • 2D/3D interior generation ( "shell" function ): add thickness to the spline of drawn houses.

  • Hole-On-Wall Generation: create holes of any number of polygons on another polygon, and re-mesh the results. Useful in the case when placing windows on the wall of the house, it automatically generates corresponding holes and the 3D geometry.

  • Procedural Window: generate window based on the specified size, style, and shape in real time.

  • Surface fitting: generate an arbitrary shape of terrain based on the drawn splines or imported height contour map.

  • Hand-drawn To Spline: convert hand-drawn doodles to spline representations in the software.

• 3D geometry, procedural mesh, and Computer Vision development.

Spatial Integrated Systems

Software Engineer

Develop and maintain 3D Imaging, Viewing, and Editing software. Some highlights:

• Developed a 3D viewing/editing software

• Shell function: add thickness to a surface, and make it a compact model with volume, for ease of 3D printing.

• 3D registration algorithm: for face recognition as well as fusing 3D data from different view angles into one entity.

• 3D imaging development: including structured light stereo 3D scan, passive-light stereo 3D scan, and laser scan.

• 3D calculation speed boost (close to real time): by NVIDIA parallel computation technique.

North Carolina State University

Ph.D. Student, Research and Teaching Assistant

Biomedical Image Segmentation and Registration:

• Proposed a constrained Mumford-Shah type energy functional incorporated with an information-theoretic view for brain MRI T1-Map and T1-weighted image segmentation in 2-D and 3-D.

• Proposed a joint segmentation and RF-inhomogeneity calibration method for accurate T1-Map generation

Layered Image/Video Segmentation-Registration and Tracking:

• Proposed a technique to jointly segment and register aerial layered images. The technique combined multiphase active contour method and joint segmentation-registration technique in a variational framework.

• Proposed a constrained optical flow calculation between consecutive frames for layered videos tracking.

LIDAR data processing:

• analyzing multiple view LIDAR (Light Detection And Ranging) images. Proposed two methods for determining the canonical views of 3D objects, by using Minimum Description Length Criterion and Compressive Sensing Theory.

Ph.D. in Electrical and Computer Engineering

North Carolina State University GPA: 3.98

Ph.D. Thesis:

Title: Image Segmentation/Registration: a Variational Framework for 2-D and 3-D Applications

Description: proposed novel image segmentation and registration methods, in various fields such as biomedical images, in 2D and 3D, and aerial surveillance optical and LIDAR images. Object tracking was also discussed as an extension of segmentation in videos.

M.S. in Electrical and Computer Engineering

North Carolina State University

B.S. in Electrical and Computer Engineering

National Tsing-Hua University, Taiwan