Gihyeon Min
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Gihyeon Min has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
Aram Lee
1 shared publication- Aram Lee
AI Convergence Research Section, Electronics and Telecommunications Research Institute, Honam Research Division, Gwangju 61012, Republic of Korea
1 shared publication Byung-Tak Lee
1 shared publication- Byung-Tak Lee
Honam Research Center, Electronics and Telecommunications Research Institute, Gwangju, Republic of Korea
1 shared publication HongYeon Yu
1 shared publicationHyun Seo Kang
1 shared publication
Latest Publications
- 28 February 2024
- Published by Korean Institute of Information Technology in The Journal of Korean Institute of Information Technology
- Vol. 22 (2) , 35-42
Diabetic Foot Ulcer Diagnosis using Deep Learning
당뇨성 족부궤양 질환의 자동 진단을 위해 딥러닝 전이학습을 이용한 질환 분류 모델을 개발하였다. 본 연구에서는 당뇨성 족부궤양의 공개된 이미지 데이터셋(DFU challenge 2021)을 활용하여 당뇨성 족부궤양을 정상과 질환을 분류할 수 있는 모델과 감염 및 허혈, 복합성 궤양 및 기타 궤양과 분류할 수 있는 자동화된 의료 이미지 질환 진단 모델을 구현하였다. 이를 위해 6가지 심층 신경망 모델을 사용하여 학습하였고 모델의 정확도, 민감도, 정밀도, f1-score 성능을 모델별로 비교 평가하여 당뇨성 족부궤양 분류에 적합한 성능의 모델을 제시하였다. 모델의 성능을 개선하기 위해 데이터 증강 및 불균형 데이터 처리를 활용하였고, 그 중 EfficientNetB3 모델을 활용하여 당뇨성 족부궤양 데이터에 대해 정확도 89%, 민감도 89%를 달성하였다.
- 28 February 2024
- Published by Korean Institute of Information Technology in The Journal of Korean Institute of Information Technology
- Vol. 22 (2) , 35-42
Diabetic Foot Ulcer Diagnosis using Deep Learning
당뇨성 족부궤양 질환의 자동 진단을 위해 딥러닝 전이학습을 이용한 질환 분류 모델을 개발하였다. 본 연구에서는 당뇨성 족부궤양의 공개된 이미지 데이터셋(DFU challenge 2021)을 활용하여 당뇨성 족부궤양을 정상과 질환을 분류할 수 있는 모델과 감염 및 허혈, 복합성 궤양 및 기타 궤양과 분류할 수 있는 자동화된 의료 이미지 질환 진단 모델을 구현하였다. 이를 위해 6가지 심층 신경망 모델을 사용하여 학습하였고 모델의 정확도, 민감도, 정밀도, f1-score 성능을 모델별로 비교 평가하여 당뇨성 족부궤양 분류에 적합한 성능의 모델을 제시하였다. 모델의 성능을 개선하기 위해 데이터 증강 및 불균형 데이터 처리를 활용하였고, 그 중 EfficientNetB3 모델을 활용하여 당뇨성 족부궤양 데이터에 대해 정확도 89%, 민감도 89%를 달성하였다.
- Open Access
- 27 February 2024
- Published by Elsevier in Journal of Cultural Heritage
- Vol. 67, 80-91
An algorithm of line segmentation and reading order sorting based on adjacent character detection: A post-processing of OCR for digitization of Chinese historical texts
- 27 February 2024
- Published by Elsevier in Journal of Cultural Heritage
- Vol. 67, 80-91
Open AccessAn algorithm of line segmentation and reading order sorting based on adjacent character detection: A post-processing of OCR for digitization of Chinese historical texts
- Open Access
- 30 November 2022
- Published by Digital Contents Society in Journal of Digital Contents Society
- Vol. 23 (11) , 2239-2248
Line Segmentation and Reading Order Detection Algorithm for Digitization of Korean Historical Texts
- 30 November 2022
- Published by Digital Contents Society in Journal of Digital Contents Society
- Vol. 23 (11) , 2239-2248
Open AccessLine Segmentation and Reading Order Detection Algorithm for Digitization of Korean Historical Texts
- 14 March 2018
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 29 (4) , 045403
Pattern-projected schlieren imaging method using a diffractive optics element
We propose a novel schlieren imaging method by projecting a random dot pattern, which is generated in a light source module that includes a diffractive optical element. All apparatuses are located in the source side, which leads to one-body sensor applications. This pattern is distorted by the deflections of schlieren objects such that the displacement vectors of random dots in the pixels can be obtained using the particle image velocity algorithm. The air turbulences induced by a burning candle, boiling pot, heater, and gas torch were successfully imaged, and it was shown that imaging up to a size of 0.7 m × 0.57 m is possible. An algorithm to correct the non-uniform sensitivity according to the position of a schlieren object was analytically derived. This algorithm was applied to schlieren images of lenses. Comparing the corrected versions to the original schlieren images, we showed a corrected uniform sensitivity of 14.15 times on average.
- 14 March 2018
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 29 (4) , 045403
Pattern-projected schlieren imaging method using a diffractive optics element
We propose a novel schlieren imaging method by projecting a random dot pattern, which is generated in a light source module that includes a diffractive optical element. All apparatuses are located in the source side, which leads to one-body sensor applications. This pattern is distorted by the deflections of schlieren objects such that the displacement vectors of random dots in the pixels can be obtained using the particle image velocity algorithm. The air turbulences induced by a burning candle, boiling pot, heater, and gas torch were successfully imaged, and it was shown that imaging up to a size of 0.7 m × 0.57 m is possible. An algorithm to correct the non-uniform sensitivity according to the position of a schlieren object was analytically derived. This algorithm was applied to schlieren images of lenses. Comparing the corrected versions to the original schlieren images, we showed a corrected uniform sensitivity of 14.15 times on average.
- 13 May 2016
- Published by Wiley in ETRI Journal
- Vol. 38 (4) , 599-605
Fast Holographic Image Reconstruction Using Phase Shifting-assisted Depth Detection Scheme in Optical Scanning Holography
- 13 May 2016
- Published by Wiley in ETRI Journal
- Vol. 38 (4) , 599-605
Fast Holographic Image Reconstruction Using Phase Shifting-assisted Depth Detection Scheme in Optical Scanning Holography