Junyoung Byun (변준영)
Education
2019.2 - present Ph.D (School of Electrical Engineering), Korea Advanced Institute of Science and Technology, South Korea
2017.3 - 2019.2 M.S. (School of Electrical Engineering), Korea Advanced Institute of Science and Technology, South Korea
2013.3 - 2017.2 B.S. (Dept. of Electrical & Electronic Engineering), Yonsei university, South Korea
Research Interests
Adversarial Attack / Defense
Trustworthy neural networks
Publications
International Conference
In summary: 2 CVPR, 2 WACV, 2 ICIP, 1 ACCV papers as the first author, 6 ICIP papers as the co-author
Junyoung Byun, Myung-Joon Kwon, Seungju Cho, Yoonji Kim, and Changick Kim. "Introducing Competition to Boost the Transferability of Targeted Adversarial Examples through Clean Feature Mixup." Accepted to IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2023. [paper] [supp] [code] (Adversarial Attack: An efficient and effective feature augmentation technique)
Junyoung Byun, Seungju Cho, Myung-Joon Kwon, Hee-Seon Kim, and Changick Kim. "Improving the Transferability of Targeted Adversarial Examples through Object-Based Diverse Input." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2022. [paper] [supp] [code] (Adversarial Attack: A novel 3D-object-based realistic input transformation technique)
Junyoung Byun, Hyojun Go, and Changick Kim. "On the Effectiveness of Small Input Noise for Defending Against Query-based Black-Box Attacks." Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 2022. [paper] [supp] (Adversarial Defense: The first observation on the effectiveness of small input noise against query-based adversarial attacks)
Junyoung Byun, Hyojun Go, and Changick Kim, "Geometrically Adaptive Dictionary Attack on Face Recognition." Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 2022. [paper] [supp] [code] (Adversarial Attack: The first memory-based query-efficient adversarial attack)
Junyoung Byun, Kyujin Shim, Hyojun Go, and Changick Kim. "Hidden Conditional Adversarial Attacks." IEEE International Conference on Image Processing (ICIP). 2022. (Adversarial Attack: A new type of adversarial attacks that are only activated under specific image conditions which do not require data poisoning)
Junyoung Byun, Hyojun Go, Seungju Cho, and Changick Kim. "EXPLOITING DOUBLY ADVERSARIAL EXAMPLES FOR IMPROVING ADVERSARIAL ROBUSTNESS." IEEE International Conference on Image Processing (ICIP). 2022. (Adversarial Defense: A new regularization loss for adversarial training which exploits adversarial examples of adversarial examples)
Junyoung Byun*, Kyujin Shim*, and Changick Kim. "BitNet: Learning-Based Bit-Depth Expansion." Asian Conference on Computer Vision (ACCV). Springer, Cham, 2018. (* These two authors contributed equally) [paper] [code] (Image Processing: The first deep learning-based approach on bit-depth expansion)
Seungju Cho, Junyoung Byun, Myung-Joon Kwon, Yoonji Kim, and Changick Kim. "ADVERSARIAL TRAINING WITH CHANNEL ATTENTION REGULARIZATION." IEEE International Conference on Image Processing (ICIP). 2022. (Adversarial Defense)
Seungjun Jung, Junyoung Byun, Kyujin Shim, and Changick Kim. “Understanding VQA for Negative Answers through Visual and Linguistic Inference.” IEEE International Conference on Image Processing (ICIP). 2021. (Explainable AI)
Hyojun Go, Junyoung Byun, and Changick Kim. “Rethinking Training Schedules for Verifiably Robust Networks.” IEEE International Conference on Image Processing (ICIP). 2021. (Adversarial Defense)
Hyojun Go, Junyoung Byun, Byeongjun Park, M. Choi, S. Yoo, and Changick Kim. “Fine-Grained Multi-Class Object Counting.” IEEE International Conference on Image Processing (ICIP). 2021. (Crowd Counting)
Kyujin Shim, Junyoung Byun, and Changick Kim. "Multi-Step Quantization of a Multi-Scale Network for Crowd Counting." IEEE International Conference on Image Processing (ICIP). 2020. (Crowd Counting)
Minji Son, Myung-Joon Kwon, Hee-Seon Kim, Junyoung Byun, Seungju Cho, and Changick Kim. "ADAPTIVE WARPING NETWORK FOR TRANSFERABLE ADVERSARIAL ATTACKS." IEEE International Conference on Image Processing (ICIP). 2022. (Adversarial Attack)
International Journal
G. Lee, J. Park, Junyoung Byun, J. Yang, S. Kwon, C. Kim, C. Jang, J. Sim, J. Yook, and S. Park. "Parallel Signal Processing of a Wireless Pressure-Sensing Platform Combined with Machine-Learning-Based Cognition, Inspired by the Human Somatosensory System." Advanced Materials 32.8 (2020): 1906269. (ImpactFactor=32.08, Interdisciplinary Work)
G. Lee, G. Lee, Junyoung Byun, J. Yang, C. Jang, S. Kim, H. Kim, J. Park, H. Lee, J. Yook, S. Kim, and S. Park. "Deep Learning-Based Deconvolution of Mechanical Stimuli with Ti3C2TX MXene Electromagnetic Shield Architecture via Dual-Mode Wireless Signal Variation Mechanism." ACS nano 14.9 (2020): 11962-11972. (ImpactFactor=18.02, Interdisciplinary Work)
Seunghan Yang, Hyoungseob Park, Junyoung Byun, and Changick Kim. "Robust Federated Learning with Noisy Labels." IEEE Intelligent Systems 37.2 (2022): 35-43. (Federated Learning)
Domestic Conference
변준영, 심규진, 김창익. "부분적인 스크린 영상 혼합을 통한 합성곱 신경망의 영상 인식 성능 향상." 대한전자공학회 학술대회 (2019): 970-973. (Data Augmentation)
변준영, 이혁재, 김창익. "관심 객체 분할 개선을 위한 반복적 가역 그래프 컷의 선택적 적용." 대한전자공학회 학술대회 (2017): 484-487. (Salient Object Segmentation)
심규진, 변준영, 김창익. "군중 집계를 위한 다단계 양자화 학습." 대한전자공학회 학술대회 (2019): 983-985. (Crowd Counting)
김정수, 변준영, 김창익. "전이 기반 적대적 공격 방어를 위한 신경망의 특성 지도 데이터 무작위 치환." 대한전자공학회 학술대회 (2020): 360-364. (Adversarial Defense)
고강욱, 심규진, 변준영, 김창익. "합성곱 신경망의 가중치 다변화를 통한 성능 향상." 대한전자공학회 학술대회 (2021): 340-344. (Image Recognition)
김희선, 권명준, 변준영, 김창익. "원근 변환 기법을 통한 적대적 이미지의 전이성 향상." 대한전자공학회 학술대회 (2021): 611-615. (Adversarial Attack)
최명애, 유승화, 김창익, 변준영, 고효준, 김정수, 박병준. "디지털 자연보전: AI 를 이용한 DMZ 야생동물 종 및 개체수 분석." 대한지리학회 학술대회논문집 (2020): 47-48. (Interdisciplinary Work: Eco AI)
김민범, 김희선, 변준영, 김창익. "이중 무작위 변환 기법을 통한 보편적 적대적 섭동의 전이성 향상." 대한전자공학회 학술대회 (2022): 2290-2294. (Adversarial Attack)
김희선, 변준영, 손민지, 김창익. "혼합 이미지를 통한 보편적 적대적 섭동의 공격 성공률 향상." 대한전자공학회 학술대회 (2022): 1969-1972. (Adversarial Attack)
Domestic Journal
정승준, 변준영, 김창익. "설명 가능한 인공지능 기술의 소개." 전자공학회지 46.2 (2019): 55-63. (Explainable AI)
김희선, 권명준, 변준영, 김창익. "카메라 시점 변환을 이용한 전이 기반 적대적 공격 기법." 전자공학회논문지 59.7 (2022): 53-59. (Adversarial Attack)
Honors & Awards
2019년도 대한전자공학회 추계학술대회 우수 발표 논문상
[paper] 변준영, 심규진, 김창익, "부분적인 스크린 영상 혼합을 통한 합성곱 신경망의 영상 인식 성능 향상"
2019년도 가을학기 우수조교상