Jeonghun Baek
Assistant Professor at 도쿄 대학
Tokyo, Japan
Summary
🤩
Rockstar🎓
Top SchoolJeonghun Baek is an Assistant Professor at The University of Tokyo with 11 years of experience in AI research, specializing in OCR, computer vision, and NLP. He holds a PhD from UTokyo and has blended academic rigor with industry impact through roles at NAVER, NCSOFT, and a student-researcher stint with Google's OCR team. His open-source contributions include practical fixes and optimizations to the widely used deep-text-recognition-benchmark (ICCV 2019), improving CTC loss handling and training pipelines. He has secured JSPS funding during his postdoc and collaborated on applied projects like onomatopoeia recognition for comic translation, showing a knack for bringing research into niche real-world applications. Based in Tokyo, he combines high academic grades and publication-quality research with hands-on engineering that improves reproducible OCR tooling.
11 years of coding experience
5 years of employment as a software developer
University of TokyoMaster's degree, Informatics, 3.90/4.00, Master's degree, Informatics, 3.90/4.00 at Kyoto UniversityBachelor's degree, Mechanical Systems Engineering, 3.89/4.00, Bachelor's degree, Mechanical Systems Engineering, 3.89/4.00 at Tokyo University of Agriculture and Technology
Korean, Japanese, English
Github Skills (14)
ocr10
mask-rcnn10
faster-rcnn10
pytorch10
eval10
text-recognition10
deep-learning10
trainings10
ocra10
loss10
python10
evaluation10
modeling10
ctc10
Programming languages (2)
Github contributions (5)
clovaai/deep-text-recognition-benchmark
Apr 2019 - Dec 2020
Text recognition (optical character recognition) with deep learning methods, ICCV 2019
Role in this project:
ML Engineer
ML EngineerContributions:76 commits, 18 PRs, 119 pushes in 1 year 8 months
Contributions summary:Jeonghun's contributions primarily involve modifications and updates to the text recognition model within the repository. These changes include fixing issues related to CTC loss functions, updating transformation and dataset configurations, and improving model training and evaluation procedures. They implemented adjustments to the CTC loss calculation and the overall model training process, specifically addressing issues and improving functionality. Furthermore, the user updated the best model settings, indicating efforts toward model optimization.
ocr-recognitioncrnncharacter-recognitionoptical-character-recognitiontext-recognition
ku21fan/COO-Comic-Onomatopoeia
Jul 2022 - Dec 2022
COO: Comic onomatopoeia dataset (ECCV 2022)
Contributions:78 commits, 2 PRs, 27 pushes in 5 months
comic-onomatopoeiaeccv2022link-predictiontext-detectiontext-recognition
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.