CakeResume Talent Search

Advanced filters
Bật
4-6 năm
6-10 năm
10-15 năm
Hơn 15 năm
Avatar of the user.
Avatar of the user.
Data Science Competition Participant @Self-Employed
2020 ~ Hiện tại
資料科學家
Trong vòng một tháng
Microsoft Office
python
machine learning
Đang học tập
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
4-6 năm
National University of Singapore
Department of building
Avatar of Yuchun Lai.
Avatar of Yuchun Lai.
Past
Frontend Engineering Manager, Data Science @Vpon Big Data Group
2022 ~ 2023
Frontend Engineer, Full Stack Engineer
Trong vòng một tháng
to system planning and front-end architecture decisions for security, stability, and scalability. 5. Implemented Git Flow and Github Actions for efficient team collaboration. 6. Wrote unit tests, E2E tests using Jest, Cypress, and Mocks Server for code and system stability. Sr. Frontend Engineer, Data Science • Vpon Big Data Group MayFebruary 2022 | Taipei, Taiwan 1. U sing React and TypeScript to build a large-scale data platform, featuring data visualizations and audience segments. 2. Using deck.gl and vector tiles to build geo data visualizations, with loading times under 1s
HTML
CSS
React
Thất nghiệp
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
10-15 năm
YZU University (元智大學)
Information Communication
Avatar of the user.
Avatar of the user.
Past
博士後研究員 @洛桑大學神經發育疾病實驗室
2023 ~ 2023
Data Scientist, Data Analyst, Machine Learning Engineer
Trong vòng một tháng
Data Science
Data Analysis
Machine Learning
Thất nghiệp
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
4-6 năm
洛桑聯邦理工學院(EPFL)
神經科學
Avatar of the user.
Avatar of the user.
Past
Data Engineer @Rooit Inc. (XO App)
2023 ~ 2023
AI工程師、機器學習工程師、深度學習工程師、資料科學家、Machine Learning Engineer、Deep Learning Engineer、Data Scientist
Trong vòng một tháng
Python
Data Analysis
Data Science
Thất nghiệp
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
6-10 năm
中國醫藥大學(China Medical University)
臨床醫學研究所
Avatar of 賴泳瑄.
Avatar of 賴泳瑄.
老闆 @歐趴飲料店
2012 ~ 2018
AIOT開發工程師
Trong vòng một tháng
賴泳瑄(Aimons) Software [email protected] 大家好,我在大學期間創業開設飲料店持續經營5年,曾經同時管理2間店與12位工讀生,後來因為對於資訊的興趣繼續讀碩士,在碩士期間研究並學習AI與各種語言工具配合過5個科技部計畫,畢業後在東海擔任兼
機器學習、大數據分析、邊緣運算、資料探勘
Application Development
Data Science
Đã có việc làm
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
6-10 năm
Tunghai University
資訊
Avatar of Chun-Jung Huang.
Avatar of Chun-Jung Huang.
OPC Chief Engineer @TSMC
2020 ~ Hiện tại
AI工程師、機器學習工程師、深度學習工程師、資料科學家、Machine Learning Engineer、Deep Learning Engineer、Data Scientist
Trong vòng một tháng
and reducing turnaround time. ◆Managed large-scale data processing tasks, demonstrating expertise in analyzing and handling datasets of hundreds of millions, to bolster model development and optimization. ◆Excelled in distributed computing, optimizing code execution across thousands of systems to significantly improve processing speed and efficiency. ◆Developed sophisticated data visualization tools to distill complex datasets into actionable insights, aiding strategic decision-making. The University of Tokyo, Foreign Researcher (OctSep◆Pioneered a neural network-based approach for cell image classification and data visualization, enhancing lab capabilities in biological research. ◆Designed a user-friendly GUI for neural network model
Deep learning with TensorFlow
Translational Research
Clinical Research
Đã có việc làm
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
4-6 năm
National Chiao-Tung University
Ph.D. - Clinical Engineering
Avatar of 宋浩茹 Ellie Sung.
AI工程師、機器學習工程師、深度學習工程師、資料科學家、Machine Learning Engineer、Deep Learning Engineer、Data Scientist
Trong vòng một tháng
Pandas, Hugging Face, Gradio, LangChain, Tensorflow, Keras, FastAPI, OpenCV, Airflow, Git, Docker, Jenkins, Line Bot , Azure Bot Service, Tableau ML & NLP Techniques: LMOps, RAG, Fine-tune LLMs, Text Generation, Multi-Document Summarization, Recommendation System, Text Classification, Named Entity Recognition, CoT Research and Work Experience Research Assistant OctPresent Institute of Information Science, Academia Sinica, Taiwan Natural Language and Knowledge Processing Lab (NLP Lab) National Taiwan University Hospital (NTUH): Focused on exploring the generality and expertise of LLMs ( e.g., LLaMA2, Zephyr. ) in medical Q&A applications, improving model performance by 2.01% . Efficiency Optimization: Utilized Low
Python
R
Natural Language Processing (NLP)
Đã có việc làm
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
4-6 năm
國立政治大學(National Chengchi University)
資訊科學系
Avatar of CHUN-HSIEN (TEDDY) LIN.
Avatar of CHUN-HSIEN (TEDDY) LIN.
資深管理師 @群聯電子 PHISON Electronics
2020 ~ Hiện tại
Trong vòng một tháng
CHUN-HSIEN (TEDDY) LIN E-mail: [email protected] Tel:South Dist., Taichung City, Taiwan (R.O.C.) My name is Teddy Lin, and I come from Miaoli, Taiwan. I am 39 years old and hold three master's degrees. I specialized in Industrial Management at Chung Hua University in Taiwan, Environmental Science and Management at Samford University in the United States, and Computer Science at National Chung Hsing University in Taiwan. Currently, I am pursuing a Ph.D. in Data Science and Engineering at National Yang Ming Chiao
AutoCAD
Operating Performance Analysis
Occupational Health Safety Management
Đã có việc làm
Sẵn sàng phỏng vấn
Full-time / Quan tâm đến làm việc từ xa
Hơn 15 năm
National Yang Ming Chiao Tung University
Institute of Computer Science and Engineering
Avatar of Zheng Tzer Lee (李政澤).
Avatar of Zheng Tzer Lee (李政澤).
Consultant @Startup
2023 ~ 2024
Pre-sales/PM/Business Consultant/Business Analyst/System Analyst
Trong vòng một tháng
戶觀察與深度訪談 • 透過線上線下顧客行為分析,驗證市場機會 • 個案研究與產業趨勢分析 • 整合內部資料與台灣政府的 Open data,自動化業務開發環節 • 透過 Tableau Server 建置公司資料倉庫 • 使用 Python 和 Tableau Prep 自動化 ETL,建立datapineline • 領導 A/B 測試設計與執行 Data Science
Python
Tableau Prep/Tableau Desktop
ETL
Đã có việc làm
Sẵn sàng phỏng vấn
Full-time / Không quan tâm đến làm việc từ xa
4-6 năm
Fu Jen Catholic University
Brand and Fashion Management
Avatar of moh yanni fikri.
Avatar of moh yanni fikri.
Past
Electrical Maintenance @PT. Pabrik Kertas Tjiwi Kimia Tbk.
2021 ~ 2023
Engineer
Trong vòng một tháng
in preparation for when there is a tool has problem or automation project Education Politeknik Perkapalan Negeri Surabaya Automation Engineering,GPASkill Preventive Maintenance ( Preparing weekly, monthly summary, and exception reports ) AC / DC Drive Wiring Diagram ( AUTOCAD, EPLAN P8 Electrical ) PLC and HMI Programming SCADA, Power Inverter Project and People Management Electrical Troubleshooting Computerized Maintenance Management Systems (CMMS) Data Analysis and Visualization Machine Learning Certification Data Science and Machine Learning - Purwadhika Digital Technology SchoolPurwadhikaSystem 800xA with AC 800M Hardware Maintenance and Troubleshooting - ABBInstrument Inspector level 2 - Inspector TrainingInstrument Inspector level 1 - Inspector TrainingPLC Intermediate Engineer - PPNSIndustrial Automation System Design - BNSP
Data Science
Python
Machine Learning
Thất nghiệp
Sẵn sàng phỏng vấn
Full-time / Không quan tâm đến làm việc từ xa
4-6 năm
Politeknik Perkapalan Negeri Surabaya
Automation Engineering

Kế hoạch tuyển dụng đơn giản và hiệu quả nhất

Tìm kiếm 40k+ CV và chủ động liên hệ với ứng viên để đạt hiệu quả tuyển dụng cao hơn. Sự lựa chọn của hàng trăm công ty.

  • Duyệt tất cả kết quả tìm kiếm
  • Bắt đầu trò chuyện không giới hạn
  • CV chỉ có thể truy cập bởi các công ty trả phí
  • Xem địa chỉ email và số điện thoại của người dùng
Bí quyết tìm kiếm
1
Search a precise keyword combination
senior backend php
If the number of the search result is not enough, you can remove the less important keywords
2
Use quotes to search for an exact phrase
"business development"
3
Use the minus sign to eliminate results containing certain words
UI designer -UX
Chỉ những CV công khai mới khả dụng với gói tài khoản miễn phí.
Nâng cấp lên gói nâng cao để xem tất cả các kết quả tìm kiếm bao gồm hàng chục nghìn hồ sơ xin việc độc quyền trên CakeResume.

Definition of Reputation Credits

Technical Skills
Specialized knowledge and expertise within the profession (e.g. familiar with SEO and use of related tools).
Problem-Solving
Ability to identify, analyze, and prepare solutions to problems.
Adaptability
Ability to navigate unexpected situations; and keep up with shifting priorities, projects, clients, and technology.
Communication
Ability to convey information effectively and is willing to give and receive feedback.
Time Management
Ability to prioritize tasks based on importance; and have them completed within the assigned timeline.
Teamwork
Ability to work cooperatively, communicate effectively, and anticipate each other's demands, resulting in coordinated collective action.
Leadership
Ability to coach, guide, and inspire a team to achieve a shared goal or outcome effectively.
Trong vòng hai tháng
Software Engineer
Logo of 奕瑞科技有限公司.
奕瑞科技有限公司
2022 ~ Hiện tại
Taipei City, Taiwan
Professional Background
Tình trạng hiện tại
Đã có việc làm
Tiến trình tìm việc
Tắt trạng thái tìm việc
Professions
Software Engineer, Machine Learning Engineer
Fields of Employment
Trí tuệ nhân tạo/ Máy học, An ninh mạng, Dịch vụ cung cấp thông tin
Kinh nghiệm làm việc
Dưới 1 năm
Management
Chưa có
Kỹ năng
C++
html + css + javascript
Python
Django Framework
Machine Learning
Deep learning with TensorFlow
Ngôn ngữ
Chinese
Bản địa hoặc song ngữ
English
Trung cấp
Job search preferences
Vị trí
software engineer, data engineer, machine learning engineer
Loại hình công việc
Full-time
Địa điểm
台灣台北市, 台灣新北市, 台灣新竹市, 台灣桃園市
Làm việc từ xa
Quan tâm đến làm việc từ xa
Freelance
Đúng, tôi là một freelancer nghiệp dư.
Học vấn
Trường học
文化大學
Chuyên ngành
資訊工程
In
Profile 03 00@2x

許哲偉  Tony Hsu

   Software Engineer,喜歡思考、學習各種新技術,擅於分析與結構化處理複雜問題,樂於鼓勵他人以及督促自我,能以積極樂觀的心面對一切事情。 

 自學過 Stanford 吳恩達教授的 Deep Learning 課程與作業以及 Kaggle和 Github 等網站上大量 Open Source 的知識。


Software Engineer
  新北市,TW, Tel: 0937848413
 [email protected]

Skills


程式語言-program                        機器學習-ML-1                              機器學習-ML-2

 Front-End, Template

  • JavaScript
  • Html5 
  • CSS 
  • JQuery 
  • Ajax
  • Bootstrap
 Back-End, Framework, Crawler                   
  • Python
  • C++
  • Django 
  • Fastapi
  • Flask

 Process Data Package & Skills
  • Numpy
  • Pandas
  • Matplotlib
  • EDA
 ML Package

  • Scikit-learn
  • Tensorflow
           TFRecords
           TF Data API (Pipeline)
           TF Hub
  • Keras

  




 Computer Vision Package
  • OpenCV
  • Dlib
  • Mediapipe
  • Darknet
 Training Hardware

  • GPU Tesla K8, T4 (Colab)
  • GPU P100-16GB (Kaggle)
  • TPUv3-8 128GB (Kaggle)


版本控制  - 資料庫  

  • Git / Github 
  • MySQL
  • MongoDB


系統與開發工具 

  • VSCode
  • Jupyter / Colab / Kaggle notebook
  • Raspberry Pi-3B
  • Linux - Ubuntu18.04
  • Docker
  • AWS EC2



經歷(Experience)

奕瑞科技有限公司, Software Engineer - 2022/03 ~ 2022/11

         1. 奕瑞科技的訓練資料網站:

         與 Frontend Engineer 合作開發公司內部系統,負責 Backend,以 Object Detection 需要的 Data 為主,使用 Yolo 系列算法所需的 Data labeling XML(PascalVOC) format,再將所需要的資料訊息轉成 json 儲存在 MongoDB database。用 Nosql 應對日後百萬至千萬的資料查詢。編寫資料搜尋引擎、XML 轉 json 工具、自動匯入 DB 工具、自動上傳下載工具優化,運用 Docker 部署在 Ubuntu上。


        2. Camera Integrity Check System (AI 影像辨識妥善率監控系統)」的「友達」維護案子:

        與 System Technical Supervisor, AI Engineer 負責處理公司自行研發的系統問題,了解網路架構、IP Camera 視訊串流( RTSP 協定),使用過 Clonezilla 硬碟分割備份技術,學習解決連接 483 台監控設備遇到的問題,等等。


        3. 運維「泛亞智慧工地」案子:

        了解 Face Recognition device 規格書,MQTT 通訊協定,實作過 Subscriber and Publisher 測試工具,等等。


        4. 影片訓練資料的收集與硬體 api 串接:
        協助處理「泛亞專案」影片訓練資料的收集,使用切影片 frame 程式、編寫 frame_to_time 程式,等等工具。 協助處理「華夏塑膠」專案的 IP Speaker api 串接。

        5.泰國超商」人流、物品偵測與追蹤專案:
        協助交接與練習,Detection 使用 Darknet Yolov4 Model 做訓練,Tracking 使用 FastMot 算法判斷。


Project 開發與自學 - 伯父指導 (Guide project development and Self-Study) - 2021/05 ~ 2021/12 

Project 開發
        1. 實作人臉偵測、識別 (Face Detection, Recognition):   
        偵測與辨識人臉系統,寫入 CSV 檔管理出勤人名中英文轉換 

        2. 種族分類器 (Race Classifier):   
        以 Kaggle UTKFace datasetEDA 種族辨識,存成 TFRecords 檔使用tf.data pipeline (載入資料, 預先處理, cache, map, shuffle, prefetch),建立模型 (VGG16, ResNet50, Xception, EfficientnetB5-7-L2, EfficientnetV2-m-l-xl),使用 Transfer Learningpre-trained model weights (ImageNet) or Self-Supervised learning weights (Noisy-student, ImageNet21K, ImageNet21K-ft1K)Kaggle TPU/GPU 訓練& Fine-tuningTest Top1 accuracy ≈ 85.x%。 

        3. 物件偵測 (Custom Multi-Object Detection - using YOLOv4):   
        使用 open images dataset v6 (Google Datasets) Custom 3 classes Datasets (train 90%, test 10%),以 yolov4-custom.cfg 架構 + Colab GPU 從頭訓練 2000 次,達到 mAP=91%
 
        其餘時間寫的: 
        Web Crawler:  1. Google Image  2. Unsplash 圖庫。 
        Dataset practice:  Fashion-Mnist:  best accuracy ≈ 94~95%,  Cifar10:  best accuracy ≈ 93~94%,  CNN training model:  VGG16,  ResNet34,  ResNet50,  Fine-tuning tool:  Keras-tuner . 

        Self-Study:  
        學習 ML Official API 文件、hands on ML 書籍、Open Source,看台大李弘毅 ML Youtube 教程,練習實作 Model Architecture 與運用一些 SOTA ModelSelf-Supervised Learning 技術。

Coursera Deep Learning Specialization (Self-Study) - 2019/05 ~ 2019/11 
Instructor:  Stanford's Andrew Ng 
學習課程:
        1. Neural Networks and Deep Learning 
        2. Improving Deep Neural Networks Hyperparameter tuning, Regularization and Optimization 
        3. Structuring Machine Learning Projects 
        4. Convolutional Neural Networks 
        5. Sequence Models

學歷(Education)

2018/08 ~ 2018/12

策會 - AI 人工智慧創新應用就業養成班


訓練課程

前端網頁設計、Django 後端開發、Python Data Analysis、網頁爬蟲、Machine Learning、Deep Learning、OpenCV、AWS Cloud、LineBot、Git/Github、RaspberryPi-3B、Linux(Ubuntu18.04)、MySQL

小組專題製作:

1. Fusic 音樂網站 (5人)   2. 咖啡廳 AI Service (6人)

Took extra courses: 

- Edx & Microsoft:  Logic and Computational Thinking  

- Edx & Microsoft:  Introduction to Python for Data Science  

Paragraph image 00 00@2x

2011/09 ~ 2017/01

文化大學 - 資訊工程學系 (畢業)

學習經歷:  在大學修習時期有些課程不認真,以至於延宕畢業時間。迫使我更加珍惜努力學習,而找到編程 (programing) 之樂趣。放棄與克服之間我最終選擇後者,克服它。因此,透過不斷的練習,在資料結構 (Data Structure) 的正課上獲得84分,程式實作課總平均提高到90分。


- 參與社團: 系上系籃
- 暑期工讀: PX Mart (全聯)

- TOEIC成績: 460分 (2020/10)

Paragraph image 00 00@2x

奕瑞科技 Projects


2022/03 ~ In progress

奕瑞科技的訓練資料網站 - (Internal System)

負責 Backend,以 Object Detection 需要的 Data 為主,使用 Yolo 系列算法所需的 Data labeling XML (PascalVOC) format,再將所需要的資料訊息轉成  json 儲存在 MongoDB database。用 Nosql 應對日後百萬至千萬的資料查詢。


編寫資料搜尋引擎、XML 轉 json工具、自動匯入 DB 工具、自動上傳下載工具優化,運用 Docker 部署在 Ubuntu 上。

2022/04 ~ In progress

Camera Integrity Check System (AI 影像辨識妥善率監控系統) - (Operation and Maintenance)

與 System Technical Supervisor, AI Engineer 處理運維系統問題,了解網路架構、IP Camera 視訊串流 ( RTSP 協定),使用過 Clonezilla 硬碟分割備份技術,學習解決連接 483 台監控設備遇到的問題,等等。

2022/09 ~ In progress

泛亞智慧工地 - (Operation and Maintenance)

與 System Technical Supervisor, AI Engineer 運維「泛亞智慧工地」案子,了解 Face Recognition device 規格書,MQTT 通訊協定,實作過 Subscriber and Publisher 測試工具,等等。 


協助處理「泛亞專案」影片訓練資料的收集,使用切影片 frame 程式、編寫 frame_to_time 程式,等等工具。


協助處理「華夏塑膠」專案的 IP Speaker api 串接。

2022/03 ~ 2022/04

「泰國超商」人流、物品偵測與追蹤專案

協助交接與練習,Detection 使用 Darknet Yolov4 Model 做訓練,Tracking 使用 FastMot 算法判斷。

AI Projects


2021/05 ~ 2020/12 

Custom YOLOv4 (Multi-Object Detection Project)

軟體實作:

使用 Open Images Dataset V6 (Google Datasets) 做Custom 3 classes Datasets (train: 三個類別各 1500 張 img + annotaions, test: 三個類別各 300 張img + annotaions),以 darknet yolov4-custom.cfg 架構 + Colab GPU training 1800 iterations,達到mAP=91%。

Paragraph image 02 00@2x

工具: 

Python, OpenCV, Darknet, 

Macbook Pro Camera, VSCode, 

Colab (GPU) 

 

參考資料 & Open Source: 

ScaledYOLOv4 (Github) 

https://github.com/WongKinYiu/ScaledYOLOv4 

YOLOv4: Optimal Speed and Accuracy of Object Detection 


My Github:   

Paragraph image 04 01@2x

2021/05 ~ 2020/12

種族分類器 (Race Classifier Project) 


軟體開發

Data:

Kaggle UTKFace (Open Data) 

Data Preprocess:

Python, Numpy, Pandas, Matplotlib, EDA

Build Model: 

Tensorflow, Keras 

CNN Architecture: 

VGG16, ResNet50, Xception, EfficientnetB4-5-7-L2 (SOTA), EfficientnetV2-m-l-xl (SOTA)


Skills used

1. Data-cleaning (Sklearn IsolationForest) -> not good

2. Data-Augmentation 

3. Transfer learning 

4. Learning Rate Scheduler

5. Tensorboard 

6. ImageNet pre-trained model 

7. Self-Supervised-Learning pre-trained model (Noisy Student, ImageNet21k or 21K-ft1k) 

8. Fine-tuning

9. TFRecords (protobuffer)

10. TF Data API (shuffle -> map -> batch -> prefetch)


Hardware

1. NV GPU K8, T4 (Colab) 

2. NV GPU P100-16GB (Kaggle) 

3. TPUv3-8 128GB (Kaggle)

  • TPU Skills - Convert tf.float32  to tf.bfloat16

Problem Solved: 

Training model

  • GPU Out of Memory
  • TPUv3 (Exceeded hbm capacity) 
  • Cloud VM problem

Project process: 

分析&預處理:

使用 Kaggle UTKFace 約 23708 張 Face dataset -> 做 EDA 分析 (ex: sex, age, race) -> Data cleaning -> 將資料用Sklearn train_test_split 方法切割成 train: 80%, valid: 10%, test: 10% -> 將分好的資料寫成二進位格式轉成TFRecords 檔 (能夠在訓練時快速讀取大量資料) -> 讀取大量圖片檔案並轉成 numpy 格式,遇到 I/O 問題,使用 multiprocessing 跟容器減少讀取時間跟記憶體消耗 -> 解析 TFRecords 檔使用 tf.data pipeline (載入資料, 預先處理, cache, map, shuffle, prefetch) 

建模&訓練:

建立模型 (ex: VGG16, ResNet50, Xception, EfficientnetB5-7-L2, EfficientnetV2-m-l-xl) -> 使用 Transfer Learning 加 pre-trained model weights (ex: ImageNet) or Self-Supervised learning weights (ex: Noisy-student, ImageNet21K, ImageNet21K-ft1K) -> Fine-tuning -> 使用 Kaggle TPU/GPU 訓練 -> Evaluate Accuracy -> Plot predict curves -> Confusion Matrix -> Visualize prediction images -> F1 score 

Test Top1 Accuracy: ≈ 85.x%                                                                                                                   My Github:  

2021/05 ~ 2020/12 

Face Detection and Recognition (Face Attendance Project)  


軟體實作:

Python, OpenCV, Pillow, Dlib, MediapipeFace_recognition


Paragraph image 02 00@2x
Paragraph image 03 00@2x

功能:

1. 偵測與辨識人臉系統,寫入CSV檔管理出勤 2. 人名中英文轉換

實作工具:

Macbook Pro Camera, VSCode








My Github:  

2018/8 ~ 2018/12

Automatic-Cafe (Group Project) 

Web 開發:

JavaScript, Html5, CSS, Bootstrap,

Nginx

軟體開發:

Tensorflow, Jupyter notebook, 

OpenCV, Tesseract OCR, Linux(Ubuntu18.04), Linebot

硬體 & 開源工具:

RaspberryPi-3B, Nvidia GPU 2080, LabelImg, Donkey Car & Ducky Car Framework

功能:

1. Web 顧客選位  

2. LineBot 語音點餐、拉花遊戲、滿意度調查服務  

3. 以 Donkey Car 架構為基礎訓練的送餐車

4. 用 LineBot 呈現以 RNN 做的詩詞

5. CNN 老鼠辨識器,用以解決倉儲中環境衛生問題。

6. 我的功能以下面的 Text Recognition 專題介紹。

Group of 6.

Paragraph image 02 00@2x

Text Recognition (My Project)

軟體實作:

Python, OpenCV, Tesseract OCR,

EAST pre-trained model and Ubuntu18.04.

功能:

Text Recognition 用在辨識顧客的牌子文字

參考資料 & Open Source: 

1. EAST: An Efficient and Accurate Scene Text Detector (Github)

2. PyImageSearch

My Github:    

https://github.com/tonyhsu32/AI-Cafe-with-machine-learning

My Demo:  https://www.youtube.com/channel/UC8Rz5NB_A_FCEAXJjIC8xqw


Paragraph image 00 00@2x

Paragraph image 00 00@2x

Web Crawler


Paragraph image 05 00@2x
Paragraph image 05 01@2x

2021/05 ~ 2020/12 

圖片爬蟲程式(Web Crawler)

1. Google Image Crawler

軟體實作:

Python, Selenium, urllib

2. Unsplash 圖庫 Crawler

軟體實作:

Python, Selenium, urllib, BeautifulSoup

功能: 自動化圖片抓取

My Github:  

Web Projects


Paragraph image 03 00@2x

2018/8 ~ 2018/12

Music Web (Group Project)

前端開發:

JavaScript, Html5, CSS, Bootstrap

後端開發:

Python, Django, MySQL

功能:

CRUD 服務, 註冊會員, 留言板, 聊天功能 (我), 自動匹配喜好 Youtube 音樂, FB Chatbot 服務。

UI介面: 參考 Spotify 網站

Group of 5.

My Github:  https://github.com/tonyhsu32/team4project                   

葆光系統 - POS 網站開發-Case (Project)

軟體開發:

JavaScript, Html5, CSS, Bootstrap, UI

資料: 

葆光系統 - POS 管理 Data

功能: 

POS 網站首頁動態介紹 (Self-Study期間完成)

My Github:  https://github.com/tonyhsu32/FitSoft-web


Paragraph image 00 00@2x

Paragraph image 00 00@2x

Edx x Microsoft Certificate:


  1. Logic and Computational Thinking 

  2. Introduction to Python for Data Science

  3. Microsoft Professional Orientation Front-End Web Developer

  4. Essential Math for Machine Learning Python Edition

  5. Algorithms and Data Structures

  2018.8 ~ 2019.2

Coursera Certificate:


Deep Learning Specialization  

 Instructor:  Stanford's Andrew Ng

 5 courses: 

        - Neural Networks and Deep Learning 

        - Improving Deep Neural Networks Hyperparameter tuning, Regularization and Optimization 

        - Structuring Machine Learning Projects 

        - Convolutional Neural Networks 

        - Sequence Models

             

 2019.5 ~ 2019.11    Coursera link:   

      ( Self-study )

Paragraph image 01 00@2x
Resume
Hồ sơ của tôi
Profile 03 00@2x

許哲偉  Tony Hsu

   Software Engineer,喜歡思考、學習各種新技術,擅於分析與結構化處理複雜問題,樂於鼓勵他人以及督促自我,能以積極樂觀的心面對一切事情。 

 自學過 Stanford 吳恩達教授的 Deep Learning 課程與作業以及 Kaggle和 Github 等網站上大量 Open Source 的知識。


Software Engineer
  新北市,TW, Tel: 0937848413
 [email protected]

Skills


程式語言-program                        機器學習-ML-1                              機器學習-ML-2

 Front-End, Template

  • JavaScript
  • Html5 
  • CSS 
  • JQuery 
  • Ajax
  • Bootstrap
 Back-End, Framework, Crawler                   
  • Python
  • C++
  • Django 
  • Fastapi
  • Flask

 Process Data Package & Skills
  • Numpy
  • Pandas
  • Matplotlib
  • EDA
 ML Package

  • Scikit-learn
  • Tensorflow
           TFRecords
           TF Data API (Pipeline)
           TF Hub
  • Keras

  




 Computer Vision Package
  • OpenCV
  • Dlib
  • Mediapipe
  • Darknet
 Training Hardware

  • GPU Tesla K8, T4 (Colab)
  • GPU P100-16GB (Kaggle)
  • TPUv3-8 128GB (Kaggle)


版本控制  - 資料庫  

  • Git / Github 
  • MySQL
  • MongoDB


系統與開發工具 

  • VSCode
  • Jupyter / Colab / Kaggle notebook
  • Raspberry Pi-3B
  • Linux - Ubuntu18.04
  • Docker
  • AWS EC2



經歷(Experience)

奕瑞科技有限公司, Software Engineer - 2022/03 ~ 2022/11

         1. 奕瑞科技的訓練資料網站:

         與 Frontend Engineer 合作開發公司內部系統,負責 Backend,以 Object Detection 需要的 Data 為主,使用 Yolo 系列算法所需的 Data labeling XML(PascalVOC) format,再將所需要的資料訊息轉成 json 儲存在 MongoDB database。用 Nosql 應對日後百萬至千萬的資料查詢。編寫資料搜尋引擎、XML 轉 json 工具、自動匯入 DB 工具、自動上傳下載工具優化,運用 Docker 部署在 Ubuntu上。


        2. Camera Integrity Check System (AI 影像辨識妥善率監控系統)」的「友達」維護案子:

        與 System Technical Supervisor, AI Engineer 負責處理公司自行研發的系統問題,了解網路架構、IP Camera 視訊串流( RTSP 協定),使用過 Clonezilla 硬碟分割備份技術,學習解決連接 483 台監控設備遇到的問題,等等。


        3. 運維「泛亞智慧工地」案子:

        了解 Face Recognition device 規格書,MQTT 通訊協定,實作過 Subscriber and Publisher 測試工具,等等。


        4. 影片訓練資料的收集與硬體 api 串接:
        協助處理「泛亞專案」影片訓練資料的收集,使用切影片 frame 程式、編寫 frame_to_time 程式,等等工具。 協助處理「華夏塑膠」專案的 IP Speaker api 串接。

        5.泰國超商」人流、物品偵測與追蹤專案:
        協助交接與練習,Detection 使用 Darknet Yolov4 Model 做訓練,Tracking 使用 FastMot 算法判斷。


Project 開發與自學 - 伯父指導 (Guide project development and Self-Study) - 2021/05 ~ 2021/12 

Project 開發
        1. 實作人臉偵測、識別 (Face Detection, Recognition):   
        偵測與辨識人臉系統,寫入 CSV 檔管理出勤人名中英文轉換 

        2. 種族分類器 (Race Classifier):   
        以 Kaggle UTKFace datasetEDA 種族辨識,存成 TFRecords 檔使用tf.data pipeline (載入資料, 預先處理, cache, map, shuffle, prefetch),建立模型 (VGG16, ResNet50, Xception, EfficientnetB5-7-L2, EfficientnetV2-m-l-xl),使用 Transfer Learningpre-trained model weights (ImageNet) or Self-Supervised learning weights (Noisy-student, ImageNet21K, ImageNet21K-ft1K)Kaggle TPU/GPU 訓練& Fine-tuningTest Top1 accuracy ≈ 85.x%。 

        3. 物件偵測 (Custom Multi-Object Detection - using YOLOv4):   
        使用 open images dataset v6 (Google Datasets) Custom 3 classes Datasets (train 90%, test 10%),以 yolov4-custom.cfg 架構 + Colab GPU 從頭訓練 2000 次,達到 mAP=91%
 
        其餘時間寫的: 
        Web Crawler:  1. Google Image  2. Unsplash 圖庫。 
        Dataset practice:  Fashion-Mnist:  best accuracy ≈ 94~95%,  Cifar10:  best accuracy ≈ 93~94%,  CNN training model:  VGG16,  ResNet34,  ResNet50,  Fine-tuning tool:  Keras-tuner . 

        Self-Study:  
        學習 ML Official API 文件、hands on ML 書籍、Open Source,看台大李弘毅 ML Youtube 教程,練習實作 Model Architecture 與運用一些 SOTA ModelSelf-Supervised Learning 技術。

Coursera Deep Learning Specialization (Self-Study) - 2019/05 ~ 2019/11 
Instructor:  Stanford's Andrew Ng 
學習課程:
        1. Neural Networks and Deep Learning 
        2. Improving Deep Neural Networks Hyperparameter tuning, Regularization and Optimization 
        3. Structuring Machine Learning Projects 
        4. Convolutional Neural Networks 
        5. Sequence Models

學歷(Education)

2018/08 ~ 2018/12

策會 - AI 人工智慧創新應用就業養成班


訓練課程

前端網頁設計、Django 後端開發、Python Data Analysis、網頁爬蟲、Machine Learning、Deep Learning、OpenCV、AWS Cloud、LineBot、Git/Github、RaspberryPi-3B、Linux(Ubuntu18.04)、MySQL

小組專題製作:

1. Fusic 音樂網站 (5人)   2. 咖啡廳 AI Service (6人)

Took extra courses: 

- Edx & Microsoft:  Logic and Computational Thinking  

- Edx & Microsoft:  Introduction to Python for Data Science  

Paragraph image 00 00@2x

2011/09 ~ 2017/01

文化大學 - 資訊工程學系 (畢業)

學習經歷:  在大學修習時期有些課程不認真,以至於延宕畢業時間。迫使我更加珍惜努力學習,而找到編程 (programing) 之樂趣。放棄與克服之間我最終選擇後者,克服它。因此,透過不斷的練習,在資料結構 (Data Structure) 的正課上獲得84分,程式實作課總平均提高到90分。


- 參與社團: 系上系籃
- 暑期工讀: PX Mart (全聯)

- TOEIC成績: 460分 (2020/10)

Paragraph image 00 00@2x

奕瑞科技 Projects


2022/03 ~ In progress

奕瑞科技的訓練資料網站 - (Internal System)

負責 Backend,以 Object Detection 需要的 Data 為主,使用 Yolo 系列算法所需的 Data labeling XML (PascalVOC) format,再將所需要的資料訊息轉成  json 儲存在 MongoDB database。用 Nosql 應對日後百萬至千萬的資料查詢。


編寫資料搜尋引擎、XML 轉 json工具、自動匯入 DB 工具、自動上傳下載工具優化,運用 Docker 部署在 Ubuntu 上。

2022/04 ~ In progress

Camera Integrity Check System (AI 影像辨識妥善率監控系統) - (Operation and Maintenance)

與 System Technical Supervisor, AI Engineer 處理運維系統問題,了解網路架構、IP Camera 視訊串流 ( RTSP 協定),使用過 Clonezilla 硬碟分割備份技術,學習解決連接 483 台監控設備遇到的問題,等等。

2022/09 ~ In progress

泛亞智慧工地 - (Operation and Maintenance)

與 System Technical Supervisor, AI Engineer 運維「泛亞智慧工地」案子,了解 Face Recognition device 規格書,MQTT 通訊協定,實作過 Subscriber and Publisher 測試工具,等等。 


協助處理「泛亞專案」影片訓練資料的收集,使用切影片 frame 程式、編寫 frame_to_time 程式,等等工具。


協助處理「華夏塑膠」專案的 IP Speaker api 串接。

2022/03 ~ 2022/04

「泰國超商」人流、物品偵測與追蹤專案

協助交接與練習,Detection 使用 Darknet Yolov4 Model 做訓練,Tracking 使用 FastMot 算法判斷。

AI Projects


2021/05 ~ 2020/12 

Custom YOLOv4 (Multi-Object Detection Project)

軟體實作:

使用 Open Images Dataset V6 (Google Datasets) 做Custom 3 classes Datasets (train: 三個類別各 1500 張 img + annotaions, test: 三個類別各 300 張img + annotaions),以 darknet yolov4-custom.cfg 架構 + Colab GPU training 1800 iterations,達到mAP=91%。

Paragraph image 02 00@2x

工具: 

Python, OpenCV, Darknet, 

Macbook Pro Camera, VSCode, 

Colab (GPU) 

 

參考資料 & Open Source: 

ScaledYOLOv4 (Github) 

https://github.com/WongKinYiu/ScaledYOLOv4 

YOLOv4: Optimal Speed and Accuracy of Object Detection 


My Github:   

Paragraph image 04 01@2x

2021/05 ~ 2020/12

種族分類器 (Race Classifier Project) 


軟體開發

Data:

Kaggle UTKFace (Open Data) 

Data Preprocess:

Python, Numpy, Pandas, Matplotlib, EDA

Build Model: 

Tensorflow, Keras 

CNN Architecture: 

VGG16, ResNet50, Xception, EfficientnetB4-5-7-L2 (SOTA), EfficientnetV2-m-l-xl (SOTA)


Skills used

1. Data-cleaning (Sklearn IsolationForest) -> not good

2. Data-Augmentation 

3. Transfer learning 

4. Learning Rate Scheduler

5. Tensorboard 

6. ImageNet pre-trained model 

7. Self-Supervised-Learning pre-trained model (Noisy Student, ImageNet21k or 21K-ft1k) 

8. Fine-tuning

9. TFRecords (protobuffer)

10. TF Data API (shuffle -> map -> batch -> prefetch)


Hardware

1. NV GPU K8, T4 (Colab) 

2. NV GPU P100-16GB (Kaggle) 

3. TPUv3-8 128GB (Kaggle)

  • TPU Skills - Convert tf.float32  to tf.bfloat16

Problem Solved: 

Training model

  • GPU Out of Memory
  • TPUv3 (Exceeded hbm capacity) 
  • Cloud VM problem

Project process: 

分析&預處理:

使用 Kaggle UTKFace 約 23708 張 Face dataset -> 做 EDA 分析 (ex: sex, age, race) -> Data cleaning -> 將資料用Sklearn train_test_split 方法切割成 train: 80%, valid: 10%, test: 10% -> 將分好的資料寫成二進位格式轉成TFRecords 檔 (能夠在訓練時快速讀取大量資料) -> 讀取大量圖片檔案並轉成 numpy 格式,遇到 I/O 問題,使用 multiprocessing 跟容器減少讀取時間跟記憶體消耗 -> 解析 TFRecords 檔使用 tf.data pipeline (載入資料, 預先處理, cache, map, shuffle, prefetch) 

建模&訓練:

建立模型 (ex: VGG16, ResNet50, Xception, EfficientnetB5-7-L2, EfficientnetV2-m-l-xl) -> 使用 Transfer Learning 加 pre-trained model weights (ex: ImageNet) or Self-Supervised learning weights (ex: Noisy-student, ImageNet21K, ImageNet21K-ft1K) -> Fine-tuning -> 使用 Kaggle TPU/GPU 訓練 -> Evaluate Accuracy -> Plot predict curves -> Confusion Matrix -> Visualize prediction images -> F1 score 

Test Top1 Accuracy: ≈ 85.x%                                                                                                                   My Github:  

2021/05 ~ 2020/12 

Face Detection and Recognition (Face Attendance Project)  


軟體實作:

Python, OpenCV, Pillow, Dlib, MediapipeFace_recognition


Paragraph image 02 00@2x
Paragraph image 03 00@2x

功能:

1. 偵測與辨識人臉系統,寫入CSV檔管理出勤 2. 人名中英文轉換

實作工具:

Macbook Pro Camera, VSCode








My Github:  

2018/8 ~ 2018/12

Automatic-Cafe (Group Project) 

Web 開發:

JavaScript, Html5, CSS, Bootstrap,

Nginx

軟體開發:

Tensorflow, Jupyter notebook, 

OpenCV, Tesseract OCR, Linux(Ubuntu18.04), Linebot

硬體 & 開源工具:

RaspberryPi-3B, Nvidia GPU 2080, LabelImg, Donkey Car & Ducky Car Framework

功能:

1. Web 顧客選位  

2. LineBot 語音點餐、拉花遊戲、滿意度調查服務  

3. 以 Donkey Car 架構為基礎訓練的送餐車

4. 用 LineBot 呈現以 RNN 做的詩詞

5. CNN 老鼠辨識器,用以解決倉儲中環境衛生問題。

6. 我的功能以下面的 Text Recognition 專題介紹。

Group of 6.

Paragraph image 02 00@2x

Text Recognition (My Project)

軟體實作:

Python, OpenCV, Tesseract OCR,

EAST pre-trained model and Ubuntu18.04.

功能:

Text Recognition 用在辨識顧客的牌子文字

參考資料 & Open Source: 

1. EAST: An Efficient and Accurate Scene Text Detector (Github)

2. PyImageSearch

My Github:    

https://github.com/tonyhsu32/AI-Cafe-with-machine-learning

My Demo:  https://www.youtube.com/channel/UC8Rz5NB_A_FCEAXJjIC8xqw


Paragraph image 00 00@2x

Paragraph image 00 00@2x

Web Crawler


Paragraph image 05 00@2x
Paragraph image 05 01@2x

2021/05 ~ 2020/12 

圖片爬蟲程式(Web Crawler)

1. Google Image Crawler

軟體實作:

Python, Selenium, urllib

2. Unsplash 圖庫 Crawler

軟體實作:

Python, Selenium, urllib, BeautifulSoup

功能: 自動化圖片抓取

My Github:  

Web Projects


Paragraph image 03 00@2x

2018/8 ~ 2018/12

Music Web (Group Project)

前端開發:

JavaScript, Html5, CSS, Bootstrap

後端開發:

Python, Django, MySQL

功能:

CRUD 服務, 註冊會員, 留言板, 聊天功能 (我), 自動匹配喜好 Youtube 音樂, FB Chatbot 服務。

UI介面: 參考 Spotify 網站

Group of 5.

My Github:  https://github.com/tonyhsu32/team4project                   

葆光系統 - POS 網站開發-Case (Project)

軟體開發:

JavaScript, Html5, CSS, Bootstrap, UI

資料: 

葆光系統 - POS 管理 Data

功能: 

POS 網站首頁動態介紹 (Self-Study期間完成)

My Github:  https://github.com/tonyhsu32/FitSoft-web


Paragraph image 00 00@2x

Paragraph image 00 00@2x

Edx x Microsoft Certificate:


  1. Logic and Computational Thinking 

  2. Introduction to Python for Data Science

  3. Microsoft Professional Orientation Front-End Web Developer

  4. Essential Math for Machine Learning Python Edition

  5. Algorithms and Data Structures

  2018.8 ~ 2019.2

Coursera Certificate:


Deep Learning Specialization  

 Instructor:  Stanford's Andrew Ng

 5 courses: 

        - Neural Networks and Deep Learning 

        - Improving Deep Neural Networks Hyperparameter tuning, Regularization and Optimization 

        - Structuring Machine Learning Projects 

        - Convolutional Neural Networks 

        - Sequence Models

             

 2019.5 ~ 2019.11    Coursera link:   

      ( Self-study )

Paragraph image 01 00@2x