Close

Hyeok Joon Kweon

AI Researcher

Download My CV Download My PORTFOLIO

About Me


· I am Hyeok Joon Kweon.
· B.S. in Electronics Engineering, Chungnam National University (2014.03 ~ 2021.02)
· M.S in CVIP Lab under the supervision of Prof. Donghyeon Cho of Electronics Engineering,
  Chungnam National University (2021.02 ~ 2023.02)
· Email : cjswpwjrdls@gmail.com, jjun@o.cnu.ac.kr
· My blog URL : https://velog.io/@with_jjun

Research interests


· Surveillance System
  - Image Steganography
  - Image Stitching
  - Person Re-identification
· Sensor Fusion
  - Sensor Time Synchronization
  - Sensor Calibration

Publications


· Cloth-Changing Person Re-Identification With Noisy Patch Filtering, Kweon, Hyeok-Joon, and Donghyeon Cho, IEEE Signal Processing Letters (SPL), 2023.
· Panorama Image Stitching Using Sythetic Fisheye Image, Hyeokjoon Kweon, Donghyeon Cho, Journal of the Korea Broadcasting Engineering Association, 2022.
· Deep Multi-Image Steganography with Private Keys, Hyeok-Joon Kweon*, Jinsun Park*, Sanghyun Woo and Donghyeon Cho, Electronics, 2021.

Education

Yuseong High School

Feb 2011 - Feb 2014

Science, Technology, Engineering, Math

34181 Yuseong-daero 654beon-gil, Yuseong-gu, Daejeon, 130 (Guam-dong).

Chungnam National University

Feb 2014 - Feb 2021

B.S. in Electronics Engineering

99 Daehak-ro Yuseong-gu Dajeon 34134, Republic of Korea.

Chungnam National University

Feb 2021 - Feb 2023

M.S. in Electronics Engineering

99 Daehak-ro Yuseong-gu Dajeon 34134, Republic of Korea.

Skills


Publications

Cloth-Changing Person Re-Identification with Noisy Patch Filtering

In this paper, we propose the cloth-changing ReID model that balances the cloth-related identity information and the cloth-unrelated identity information via two stream models. The first stream is the baseline ReID network that extracts features from the full RGB image. The second stream is the cloth-unrelated ReID network that only takes features from sub-patches containing head and leg regions. These two networks are trained separately, and fused during the inference stage. In addition, to reduce noisy patches during the training, we propose a noise patch filtering module (NPFM). As a result, our proposed method performs consistently better than existing ReID methods for both the same and cloth-changing settings.

View Publication

Panorama Image Stitching Using Sythetic Fisheye Image

본 논문에서는 완전한 지도학습이 가능한 가상환경을 이용하여 실제 환경과 유사한 다양한 날씨, 다양한 물체가 존재하는 가상환경에서의 카메라 세팅을 제안하였다. 카메라 센터가 각각 다른 3개의 360˚ 큐브 맵 영상을 시뮬레이터를 통해 만들었다. 만든 3개의 360˚ 큐브 맵 이미지를 fisheye projection을 통해 그림1에서 보이듯이 3개의 185˚의 시야각을 가지는 fisheye 영상을 제작하였다. 마찬가지로 3개의 카메라 센터의 중심에 정답 영상인 360˚ 파노라마 이미지를 제작하였다. 제작한 fisheye 영상과 정답 영상으로 본 논문은 멀티 호모그래피를 추정하여 다양한 깊이에 따른 이미지 스티칭이 가능하고 완전한 지도학습을 통해 정확한 파노라마 영상을 만드는 모델을 제안하였다.

View Publication

Deep Multi-Image Steganography with Private Keys

We propose deep multi-image steganography with private keys. Recently, several deep CNN-based algorithms have been proposed to hide multiple secret images in a single cover image. However, conventional methods are prone to the leakage of secret information because they do not provide access to an individual secret image and often decrypt the entire hidden information all at once. To tackle the problem, we introduce the concept of private keys for secret images. Our method conceals multiple secret images in a single cover image and generates a visually similar container image containing encrypted secret information inside. In addition, private keys corresponding to each secret image are generated simultaneously. Each private key provides access to only a single secret image while keeping the other hidden images and private keys unrevealed.

View Publication

Projects

가상 시뮬레이터를 이용한 fish eye image 제작 및 360˚ 파노라마 이미지 제작 모델 고안

우리는 완전한 정답영상을 만들 수 있는 자율주행 연구를 위한 오픈소스인 칼라 시뮬레이터를 이용하여 카메라 센터가 다른 3방향의 fisheye 영상을 만들어 실제 환경과 유사한 카메라 환경을 조성하였다. 조성한 3개의 fisheye 영상 3개를 입력을 완전한 지도학습으로 360 파노라마 영상을 가능하게 하는 이미지 스티칭 모델을 제안하였다. 최종 실험결과로는 실제 환경과 비슷한 가상의 다양한 환경과 큰 시차에도 강인한 스티칭 결과를 검증하고 현실에서도 적용 가능성을 보여준다.

View Project[1], View Project[2]

Experience


CGV 미소지기

10 Doan Central Plaza, Gyeryong-ro 132beon-gil, Yuseong-gu, Daejeon.


LIKELIKE Barista

98 Cultural Center-ro, Bongmyeong-dong, Yuseong-gu, Daejeon.

Get in Touch