Wayne Yu
Machine Learning Engineer at Authme- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
Topline Score
Bio
Experience
-
Authme
-
Taiwan
-
IT Services and IT Consulting
-
1 - 100 Employee
-
Machine Learning Engineer
-
Jun 2022 - Present
- Conduct a comprehensive study of ISO/IEC 30107-3 Presentation Attack Detection- Build a lightweight (<5MB) facial landmark model that could also predict landmark occlusion and face pose in one model- build lightweight MRZ detection and recognition model for passport and Taiwan resident card service using CNNs and Vision Transformers- Apply model quantization with QAT to shrink the model size and increase 60% inference time on mobile phone- build an MVP about large-scale 1:N facial recognitions to accelerate vector search with Facebook Faiss- Improve Card Anti-Fraud Detection models to detect anti-fraud features on Taiwan ID Card Show less
-
-
Computer Vision Intern
-
Nov 2021 - Jun 2022
- Refactor internal computer vision libraries and add unit tests with pytest for them- Increased overall code coverage from 32% to 95%- Attack face and card anti-spoofing models to improve models and acquire potential customers- Generate an eye-covered dataset for training a robust face recognition model with OpenCV- Explore how to make Document-Question-Answering model become a real product
-
-
-
AICS
-
Taiwan
-
IT Services and IT Consulting
-
1 - 100 Employee
-
Machine learning Summer Intern
-
Jul 2020 - Sep 2020
- Cooperated with hospitals to create their smart health system and in charge of Object Detection part - Cooperated with hospitals to create their smart health system and in charge of Object Detection part
-
-
-
National Cheng Kung University
-
Taiwan
-
Higher Education
-
700 & Above Employee
-
Research Assistant
-
Sep 2019 - Sep 2020
Build an Intelligent Plantar Disease Recognition mobile App to recognize plantar fasciitis- Provide 100+ patients in National Cheng Kung University Hospital with clearer information about their potential disease after treatment- Prototype is selected as College Student Research of MOST ( Ministry of Science and Technology )Deal with time series prediction problem of trading volume 196 banks with a deep learning model- Reduced the need for bankers who manage trading volume
-
-
Research Assistant
-
Sep 2018 - Sep 2019
- Recognize traffic lights and cars in Tainan with object detection model- Reduce holiday traffic jams of Tainan by 11%~15% in experimental regions
-
-
Education
-
國立成功大學
Bachelor's degree, Industrial and information management