洪柏安

Po - An Hung

  Taoyuan, Taoyuan District, Taoyuan City, Taiwan

Active and have a good self-learning ability.

Easy-going and a good listener.

Abundant curiosity.


Telephone : 0956308980

Email : [email protected]

LinkedIn : https://www.linkedin.com/in/po-an-hung-127949188/

  Skills

Programming

  • Python
  • Java
  • C++
  • Javascript

Backend

  • Spring Boot
  • Servlet

Frontend

  • Vue.js
  • Node.js
  • jQuery
  • Ajax


Container

  • Docker

Image Processing

  • Opencv

Deep Learning

  • Tensorflow
  • Pytorch

Version Control

  • Git
  • Github

  Work Experience

Intern  •  Dragonfly UAS

Jul 2021 - Dec 2021

Develop a camera module that can detect and analyze the current environment in real time, and shoot after adjusting camera parameters.

Intern  •  Industrial Technology Research Institute (ITRI)

Mar 2020 - Aug 2020

Study the performance when using Brain Float 16 data type in the Deep learning model.

  Autobiography

My name is Hong Po An, I grew up in Taoyuan, and I studied at Yuan Ze University with a master's degree near my home. Interests include playing an instrument, reading, and traveling. Going out for a walk increases knowledge and enjoys life.

I think my personality traits are three: 1. Easy-going and a good listener 2. Active and have good self-learning ability 3. Abundant curiosity. First of all, for the part that is easy-going and good at listening, I will stand from other people's perspective and think in a different position at appropriate times. I am willing to accept speeches that I want to disagree with, and check back to see if my thoughts are not comprehensive. For example, during the internship, when I, a program major, and an intern majoring in drones were discussing the issue of program integration, when they disagreed because of different starting points, I was able to listen to the two sides in a short period of time to understand the confusion and lack of consideration. Finally, after reinterpreting, a more perfect solution is obtained. Therefore, in the past, teamwork, I could achieve the goal with good communication and efficiency.

Then, in the second personality trait,  active and having a good self-learning ability, in the past student status, the place where I learned the most knowledge was not during class time but self-study time after class. I would find information on my own about the knowledge I am interested in, and discussions with professors in the field when bottlenecks are encountered. For example, during the closing hours of some classes, I always listen to the questions that other students left to ask the professor, and leave them until the last one to leave, because I hope to have the opportunity to use these questions to make up for the things I didn't notice in the past. Therefore, during the bachelor and master's periods, parents and teachers did not worry about the academic part.

Finally, in the third personality trait, in the part of rich curiosity, since I was a child, I have found that I am curious about anything I don’t understand, so I have the courage to try challenges in different fields and enjoy the process of learning. For example, on the project of the past course, although it was only a ticket to score the final grade, I tried to extend or study it as much as possible, and did not limit the topic to a tool for scoring, so in the end, I often recommended teaching professors to participate in competitions of all grades.

With my experience and profession, I believe I am ready for this position, and I will definitely develop myself to keep pace with the changes of the era in this field.

  Education

2020 - 2022

Yuan Ze University

Master of Electrical Engineering

2016 - 2020

Yuan Ze University

Bachelor of Electrical Engineering

  Proof of other competencies

Assist the school to optimize the thermal imager module

The team developed the "YZU-ECE Remote Body Temperature Monitor" in a short period of time, adding an algorithm for judging the temperature of the "head" area, so that the area with the highest temperature is limited to the head area. Temperature monitors have just got smarter.

https://ctee.com.tw/industrynews/cooperation/238368.html

Participate in Alibaba's AI competition

In 2019, I participated in the AI competition organized by Alibaba. The goal was to design a deep learning model to automatically detect defects in bottle body and bottle cap images.

In the end, the team ranked 142 out of nearly 3,000 teams around the world.

https://flic.kr/p/2mYme4o

Course Project - Mango Level Classification

Because of lack of training data, we tried some different data augmentation methods(ex: use GAN to generate data, use the image which is generated from Mask-RCNN, removing the background of the picture), and then tried some up-to-date classification models.

https://flic.kr/p/2mYpoXH

Course Project - DL Detection-Based for Industrial Inspection of PCB 
Manufacturers. 

We propose a fully automatic pipeline for defect detection, which allows users to upload PCB images in a web-based front-end implemented in Ajax, and send them to a back-end implemented in Flask for the deep learning model in the back-end to perform defect detection. To realize the opportunity to improve speed and accuracy, we detect defects through RetinaNet's efficient single-stage detection process, and transmit defective  areas back to the front end for inspection.

https://flic.kr/p/2mYfNUF

Course Project - Simple Hand Recognition-Assisted Teaching Tool

We use the traditional image process method to detect and recognize the movement of a user's hand to control the mouse movement(ex: move up, click), and control the slides(ex: next slide, previous slide).

https://flic.kr/p/2mYkCoz

Course Project - Countdown timer - prevent bedridness

We use arduino and buzzer to build an alarm clock which can play music(ex: Canon) for alarm users. After the user wants to lie for a while, he/she can press the buttons(some solutions to waiting time) on the alarm clock to let it alarm after waiting time, and the music will become more quickly and more higher pitch.

https://flic.kr/p/2mYnrDc

Course Project - OCR application

Use Ajax to implement a simple front-end web page, allowing users to upload the pictures they want to parse the text to the back-end after logging in through their Google account. Use Servlet to implement the backend containing CRUD and connect to the MySQL database. After the backend receives the POST Request from the frontend, the photo is stored for subsequent use. In addition, use Flask to implement a microservice dedicated to character recognition, and return the recognized text to the front end.

  Paper published

1. 洪柏安、曾俊雄(2019)。階層式演算法影像情緒識別及應用。TCSE2019。

2. 高偉城、李俊毅、洪柏安、林智揚、林政勳、陳仁和(2019)。基於Google Edge TPU之即時人臉辨識系統。中國機械工程學會第三十六屆全國學術研討會論文集。

3. 洪柏安、曾俊雄(2021)。經驗數據化感測模組管理器及圖表分析工具之實作。