- Kor: ํฌํธํด๋ฆฌ์ค
- Eng: Portfolio
-
2021.01 ~ 2021.02
- Encouragement Prize at the 1st Big Data/AI College Student Contest for Digital Innovation in the Shipbuilding/Maritime Industry
-
2022.07 ~ 2022.07
- Completed the SKKU-KISTI HPC:AI Summer School
-
2022.12 ~ 2023.01
- Monthly Dacon Machine Failure Diagnosis AI Contest TOP 4%
-
2023.05 ~ 2023.05
- Encouragement Prize at the Korea Economic Daily Intelligent Information SW Idea Contest.
-
2023.05 ~ 2023.06
- Grand Prize at the 2nd Performance Presentation of the Korea University Intelligent Information SW Academy 2023 (Awarded by the Director of the Information and Communication Planning Evaluation Institute)
-
2023.03 ~ 2023.06
- Completed the 2nd Term of the Korea University Intelligent Information SW Academy 2023.
-
2023.07 ~ 2023.11
- Excellence Prize at The 8th Precision Engineering (Hyper-Scale Artificial Intelligence and Smart & Green Precision Engineering Technology)
-
2023.07 ~ 2023.12
- Undergraduate researcher at SDML (Professor Sangwon Lee's Lab)
-
2023.12 ~ 2024.02
- Rehabilitation-Biomechatronics Research Lab (Professor Jonghyun Kim's Lab) Co-op
-
2024.03 ~ 2024.06
- Undergraduate researcher at RISE Lab (Professor Hyungpil Moon's Lab)
Project 1: HiMSEN Engine Abnormality Detection Analysis
- Based on the knowledge from the courses taken, preprocess the HiMSEN engine fault data, separate the data by mode and by four different systems, analyze it, and implement a fault diagnosis and cause system algorithm using ANN
- Role: Understanding Data Characteristics (Temperature, Pressure)
Data Preprocessing
- Participated in a project individually, extracted statistical features from sound in both time and frequency domains, preprocessed the data after separating it by mode (0,2), and implemented a machine fault diagnosis algorithm by ensembling IF (Isolation Forest), OCSVM (One-Class SVM), and AE (AutoEncoder)
Project 3: TOC in CCTV
- A project implemented as a multi-stage model of Tracking + Super-Resolution + image-captioning with the goal of text conversion of video. Used Yolo4Deepsort (Yolo4), SwinIR (Swin Transformer), and BLIP (vit-encoder + cross attention + LM-decoder) respectively.
- Role: Super-resolution (select model) Image-captioning (select model and finetuning)
Project 4: Development of an FDM process quality prediction model based on ANN and transfer learning for material response
- Participated in the 8th Precision Engineering Creative Competition as an undergraduate researcher in Professor Sangwon Lee's laboratory, along with graduate students. Developed a quality prediction model for the FDM process for composite materials (ABS, PLA, PETG) based on ANN and transfer learning
- Role: Data collection and preprocessing
Quality Prediction modeling
Make poster and presentation
Project 5: Vision inspection system for clothing fabric based on image object segmentation algorithm
- Participated in a system development project to inspect fabric defects (stain, hole, dyeing) using a vision camera and deep learning as an undergraduate research student. Achieved Acc 95.31%, IOU 0.902, and Inference time 54FPS with the deep learning image object segmentation model U-net++ and ensemble and threshold algorithms
- Role: Paper review
Code review
Image Data Collection
Data Labeling and Preprocessing
Project 6: Development of an anomaly detection algorithm based on multimodal learning for CNC tool wear recognition and monitoring
- For a capstone project, developed a CNC tool wear recognition and replacement notification algorithm by applying an ensemble model based on unsupervised and supervised learning, and a Rule-based model in stages. Unsupervised learning used clustering (Agglomerative method), ensemble (with anomaly detection and sampling techniques applied), a CNN classification model based on STFT images, and a Rule-based mode
- Role: Paper review (RUL and Anomaly detection)
Code review (DAMP algorithm)
EDA (t-test, MFCC use)
Data Labeling and Preprocessing
Model Selection
Sampling Method (over and undersampling)
Model train and test
- As part of a CO-OP project, carried out a normal person's Gait simulation using the gait simulation program SCONE and reinforcement learning algorithms. The DEP(controller)-MPO(reinforcement learning) algorithm was used, and a PD controller was utilized to maintain the balance of the upper body. Additionally, by modifying the walking model to automatically maintain balance, performance was further improved
- Role: Paper and Code review Coding and model Customize Model Selection Customize reward function and pd controller
- Email: [email protected]
- LinkedIn: Link