鄭凱元 (Kai-Yuan Cheng)
Machine Learning Engineer focusing on super resolution, object detection, and face recognition.
Strongly interested in deep learning and computer vision.
Graduated from National Taiwan University with Master of Mechanical Engineering.
AI/ML Engineer
Date of Birth: 1991/8/25
Mobile: +886(0)919-165-221
E-mail: [email protected]
LinkedIn: www.linkedin.com/in/m06800825
PROFESSIONAL EXPERIENCE
Data Scientist (Computer Vision), TSMC, 2022/4 - Present
- Super resolution for wafer patch enhancement
- Computer vision algorithm development.
Machine Learning Engineer (Computer Vision), VIVOTEK, 2019/2 - 2022/4
- Neural Network architecture design and optimization.
- Computer vision algorithm development.
- Design and develop steering wheel and airbag.
SKILLS
Programming
- Python
- C++
Development Tools
- PyTorch
- OpenCV
- Linux
- Docker
- Git
Domain knowledge
- Super resolurion
- Object detection
- Face detection and recognition
- Machine Learning
- Web Crawling
EDUCATION
INSTITUTE FOR INFORMATION INDUSTRY
Data Engineering and Data Analytics Intensive Training Program (2018/6 - 2018/11)
NATIONAL TAIWAN UNIVERSITY (NTU)
Master of Mechanical Engineering (2013/9 - 2015/7)
NATIONAL TAIWAN UNIVERSITY (NTU)
Bachelor of Mechanical Engineering (2009/9 - 2013/6)
CERTIFICATES
Microsoft Professional Program:
Data Science Certificate
TOEIC Golden Certificate
(915/990)
PROJECTS
Super Resolution for wafer defect detection
- Develop model to enhance wafer images for improving defect capture rate
- Improve defect capture rate by 20%
Object detection in surveillance system
- Develop models based on YoloX, Yolov4 and RetinaNet for wall mount and ceiling mount cameras.
- Model compression and deployment on camera: (a) Reduce model size by 4x (b) Reduce inference time by 2x
Pedestrian and vehicle attribute recognition
- Detect age, gender, accessories, and dress color of pedestrians. The valation accuracy is around 90%.
- Detect color of vehicles. The validation accuracy is around 92%.
Face detection and recognition
- Develop detection model based on RetinaFace
- Face alignment by using face landmarks from face detection.
- Develop recognition model based on ArcFace
- Serve as team leader, assign tasks and manage project progress.
- In charge of training data preparation and model architecture design.
- Predict China mobile users' characteristics based on their app usage, location, and device properties.
SUMMARY
- Continuously study SOTA deep learning algorithms of computer vision.
- Design Neural Network architecture and integrate into products of the company.
