莊才賢,Cai-Xian(John) Zhuang
[email protected]
+886-983-865-151
About Me
- I graduated from the master of AI group in National Taiwan University of Science and Technology. I like to use AI models to solve some problems that especially focus on computer vision and some image processing.
- I like to challenge the impossible, and I am curious about new knowledge.
- I am looking for a software engineer for a company where I can apply my abilities to help my team to achieve what clients want.
Education & Experience
National Taiwan University of Science and Technology, M.S., Electrical Engineering- AI, Jul. 2019 - Sep. 2021
- Courses included: Machine Learning, Algorithm, Deep Learning, Image and Video Processing
- Research: Video-Based Blood Pressure Estimation Using Two-Step Learning in 3D-CNN
- Predicting ages by faces achieves only 4.2 ages error.
- Predicting gender by faces achieves approximately 98% accuracy.
- Subjects included: Program Designing, Control System, Deep Learning Application Machine Learning
- Eighth place in the whole department
Skills
Language
- Python
- C++
- Arduino
- MATLAB
IDE/Editor Tool
- Jupyter Notebook
- Spyder3
- PyCharm
- Code::Blocks
- Visual Studio
- GitHub
Cloud Service & Database
- Google Colab
- MySQL
Other Knowledge
- AI (Artificial Intelligence)
- Image Preprocessing
- Algorithm
Projects
1.Video-Based Blood Pressure Estimation Using Two-Step Learning in 3D-CNN
We used only face frames to estimate blood pressure with a normal webcam. We proposed two training skills and a two-step learning method to reduce the loss. We finally achieve 6.95/6.53 mmHg (MAEs for SBP/DBP) and 7.50/6.29 (standard deviation for SBP/DBP) which pass the criteria of AAMI.
- Language : Python
- API : Pytorch, Matplotlib, Pandas, cv2, PIL
2.Voice to Sign Language Translation System
Customer says some words. Microphone will collect them. Computer will show the Sign Language films to the hearing-impaired people.
- Language : Python
- API : SpeechRecognition, jieba, Tkinter, cv2
3.The Simpsons Characters Recognition Challenge
We used Simpson dataset and CNN architecture as the model of training and prediction.
- Language : Python
- API : Keras, Matplotlib, Pandas, sklearn, cv2
4.Score Your Face
We used SCUT-FBP5500_v2 dataset and ResNet50 architecture as the model of training and score people’s faces.
- Language : Python
- API : Keras, Matplotlib, Pandas, sklearn, cv2
5.Smart Parking Lot
We combined Arduino and Python to build Smart Parking Lot. We used Arduino to be our warning light, sensors, motors and management system. Using Python to detect and identify license plate.
- Language : Arduino, Python
- API : Keras, Matplotlib, Pandas, sklearn, cv2
Contests
- Participate in 2019 11th IT help ironman competition (AI & Data)
- 2018 NTUT Senior Project Contest - The High Distinction Award
- Participate in 2017 TEL Robot Combat final contest(東京威力科創機器人大賽)
