Hello, I'm Ke-Chi Chang.

You can also call me Arc Chang.

I'm a Senior Software & AI Engineer at MediaTek Inc. for 5 years.
I received the B.S. and M.S. degree in Computer Science from National Tsing-Hua University.
I had been in Vision and Learning Lab with Prof. Hwan-Tzong Chen.
My research focuses on Computer Vision, Machine Learning, and Generative AI tasks.


Experience

2020/08 - Now

MediaTek Taiwan | Hsinchu

Senior Software & AI Engineer

Integrate the AI algorithms into low-level and high-level vision tasks.
Mainly for smartphone, tablet, and surveillance projects (OPPO/Vivo/Xiaomi/Honor).

  • AI-NR : Denoising on burst images or 1080p/2K/4K videos.
  • AI-HDR : Denoising and dynamic range compression on multi-exposure images.
  • AI-Flash : Denoising and dynamic range compression on flash and non-flash pairs.
  • AI-DM : Demosaicing on burst images.
  • AI-RM : Remosaicing on burst images.
  • AI-SR : Super resolution on single face input by generative models.
  • AI-DET : Detecting person's and animal's body/face/eye simultaneously on 30FPS videos.
  • AI-FLD : Detecting 300 facial landmarks, race, gender, and age on 30FPS face videos.
  • AI-SEG : Including semantic and instance segmentation tasks on 11 classes.
  • MediaTek ISP : Familiar with the software integration on the ISP of highend smartphone projects.

    • 2019/07 - 2020/06

    MediaTek Taiwan | Hsinchu

    Project Intern

    During my internship at MediaTek, I focused on real camera noise modeling using generative AI techniques. This work contributed to the development of novel noise reduction methods for real-world imaging systems. The research efforts resulted in a high-impact paper titled "Learning Camera-Aware Noise Models", which was successfully published and presented at ECCV 2020.

    2018/11 - 2019/08

    Umbo Computer Vision, Taiwan | Taipei (Remote Contract)

    Project Intern | Industry-Academy Cooperation

    During my internship at Umbo Computer Vision, I worked on detecting human postures in surveillance footage using multi-person 3D pose estimation with AI methods. The project not only enhanced the accuracy of human behavior understanding from video streams but also demonstrated the potential of AI-driven pose estimation in security and smart surveillance applications.

    2018/08

    Conference on Computer Vision, Graphics, and Image Processing (CVGIP) 2018

    2nd place in ”Indoor 3D deep space construction competition based on deep learning and 2D images”.

    The Indoor 3D Deep Space Construction Competition challenges participants to reconstruct detailed 3D indoor environments using deep learning and 2D images. In our approach, we adopt the LSD-SLAM (Large-Scale Direct Monocular SLAM) method, enabling real-time scene reconstruction at 15 FPS. This allows us to efficiently generate accurate 3D spatial structures from monocular video input, providing a practical solution for indoor mapping and navigation applications.

    2018/03 - 2018/11

    Automotive Research & Testing Center (ARTC), Taiwan | Changhua (Remote Contract)

    Project Intern | Industry-Academy Cooperation

    During my project internship at the Automotive Research & Testing Center (ARTC), I conducted research on LiDAR sensors and applied SLAM and point cloud detection techniques to enable a model vehicle to achieve basic autonomous driving capabilities within a testing track, maintaining an error margin of within 1 meter. This experience involved hands-on work with sensor data processing and autonomous navigation algorithms.

    2017/09 - 2018/08

    Industrial Technology Research Institute (ITRI), Taiwan | Hsinchu

    Project Intern | Industry-Academy Cooperation

    During my project internship at the Industrial Technology Research Institute (ITRI), I participated in study groups with colleagues to discuss and exchange knowledge on SLAM (Simultaneous Localization and Mapping) technologies. Additionally, I worked on applying learning-based methods to remove sky noise from point clouds captured by sensors or SLAM systems, improving the quality and accuracy of 3D mapping data.

    2015/07 - 2015/09

    PEGATRON Intern, Taiwan | Taipei

    Summer Intern

    During my summer internship at Pegatron, I developed an Android application for an access control system integrating face recognition technology. My role involved designing and implementing key features for real-time facial verification to enhance security and user convenience.



    Skills

    Python

    95%

    PyTorch

    90%

    TensorFlow

    85%

    C/C++

    75%


    Language

    Chinese

    95%

    English

    78%

    Taiwanese

    50%


    Contact information

    • (+886) 934-335-141
    • sky1236@gapp.nthu.edu.tw

    Copyright © All rights reserved | This template is made by Colorlib