| Role | Company | Period |
|---|---|---|
| ML/DL Engineer | SK Telecom | 2024.01. ~ |
| Deep Learning Researcher (Intern) | Neurocle | 2023.01. ~ 2023.08. |
| Role | Company | Period |
|---|---|---|
| Undergraduate Researcher | Computer Vision Machine Learning Lab. (Advisor. Prof. Junseok Kwon) |
2019.08. ~ 2020.08. |
| Classification | Skills |
|---|---|
| Machine Learning |    |
| Language - Favorite |  |
| Language - Useable |  |
| ETC |    |
| Awards | Grade | Date | Note |
|---|---|---|---|
| Military Service Open SW Online Hackathon | Navy Chief of Staff Award | 2021.10. | Canary |
| Barrier-free App Development Contest | Encouragement Award | 2020.11. | Zipyak |
- Bachelor of Engineering in Computer Science & Engineering, Chung-Ang University 2018. 03. ~ 2024.02.
- GPA : 4.41/4.5
2021
- Canary: Image mosaic for military AI_APP_WEB_Canary_Canary_2021, yolov5-knowledge-distillation
- Proceed with knowledge distillation in yolov5
- Analyze problems of model and data, and improve performance over 14.9%p
- Building an administrator server for log collection and deployment of the latest models
- Some part of codes receives 23 stars at github
2019
- Zipyak: Pharmacological Information Service for the Blind link
- Use google cloud vision OCR to iOS application
- Creating Pharmacological Information Retrieval Algorithm with OCR Data
- User interviews and user-friendly app design
- Photo Filter Application – ESRGAN-pytorch, ESRGAN-pytorch, ESRGAN-pytorch
- Implementing Style transfer using cnn, CycleGAN, ESRGAN
- Deploy using pip and github to use on Django
- ESRGAN receive 75 stars at github
- Fash: Fashion classification and recommendation system
- Build classification model and deploy to iOS applications using libtorch and object c++
2018
- Hair segmentation for mobile device - github.com/wonbeomjang/mobile-hair-segmentation-pytorch
- Implementing HairMatteNet and Improving Performance
- Improve performance and reduce execution time by changing model structure and quantization
- Improve IOU 0.73%p, reduce interference time 42.9% and model size 56%
- This project receives 102 stars at github