Offliner

I program therefore I am

Zheng-Yan Wu

Github Status

About

Hi, I am Zheng-Yan Wu.
I am currently a master student in National Taiwan University.
I am currently learning SLAM, 2D & 3D Computer Vision and Reinforcement Learning.

Education

  • Taichung First Senior High School (2014 / 9 ~ 2017 / 6)
  • National Taiwan Normal University (2017 / 9 ~ 2021 / 6)
    • Main Major - Mechatronic Engineering
    • Double Major - Eletrical Engineering
  • National Taiwan University (2021 / 9 ~ Now)
    • Graduate Institute of Mechanical Engineering
    • Research Scholar at NTU Robotic Lab

Teaching

  • Teaching Assistant at the NTNU ME Computer Programming course in 2019
  • Teaching Assistant at the NTNU ME Labs of Digital Logic course in 2019
  • Teaching Assistant at the NTNU ME The Principles and Application of Sensors course in 2020

University-Industry cooperation

  • Taiwan Industrial Technology Research Institute Intelligence Robot Talent Education (2019 / 3 ~ 2019 / 11)
  • Stamping Machinery Intelligent Sensing Industry-University Project Internship (2019 / 7 ~ 2019 / 8)

Special Topic

  • A CNN-based Sleep Posture Recognition System Using a Pressure-Sensitive Mattress
  • Nonlinear Signal Prediction and Causality Analysis Based on Granger Causality with Implementation of CNN-BiLSTM Model

Language

  • Native: Mandarin, Taiwanese
  • Proficient: English
  • Basic: Japanese

Skills

  • Proficient: C, C++, Python
  • Intermediate: C#, HTML5

  • Microsoft - Light-Weight Facial Landmark Prediction Challenge

    We train a model to predict 68 2D facial landmarks in a single cropped face image with high accuracy, high efficiency, and low computational costs.

  • Point Cloud Semantic Completion

    3D point cloud object completion is to restore broken and incomplete 3D point cloud objects to their complete appearance. We use the information about the semantic parts of these broken objects to be added to effectively improve the quality of the completed object, including the integrity of the whole and the delicacy of the details.

  • Play Table Hockey with a 5-DoF Mechanical Arm

    The dynamic hockey ball is tracked through the Realsense d435i RGB-D camera. When the ball enters the area that the robot arm can reach, the inverse kinematics is solved through the ball coordinates predicted by the Kalman Filter, and finally the robot arm is moved to the position.