점진적 단계학습과 가중치 전달 기반 향상된 ACGAN
Enhanced ACGAN based on Progressive Step Training and Weight Transfer
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
AI(Artificial Intelligence)의 다양한 모델 중 생성 모델, 특히 GAN(Generative Adversarial Network) 은 이미지 처리, 밀도 추정, 스타일 전이 등 다양한 응용 분야에서 성공을 거두었다. 이러한 GAN 은 CGAN(Conditional GAN), CycleGAN, BigGAN 등의 방식으로 확장 및 개선되었지만 재난 시뮬레이션, 의료 분야, 도시 계획 등 특정 분야에서는 데이터 부족과 불안정한 학습에 의한 이미지왜곡 문제로 실제 시스템 적용에 문제가 되고 있다. 본 논문에서는 클래스 항목을 판별하는ACGAN(Auxiliary Classifier GAN) 구조를 기반으로 기존 PGGAN(Progressive Growing of GAN)의점진적 학습 방식을 활용한 새로운 점진적 단계의 학습 방법론 PST(Progressive Step Training)를제안한다. PST 모델은 기존 방법 대비 70.82% 빠른 안정화, 51.3% 낮은 표준 편차, 후반 고해상도의 안정적 손실값 수렴 그리고 94.6% 빠른 손실 감소를 달성한다.
Abstract
Among the generative models in Artificial Intelligence (AI), especially Generative Adversarial Network (GAN) has been successful in various applications such as image processing, density estimation, and style transfer. While the GAN models including Conditional GAN (CGAN), CycleGAN, BigGAN, have been extended and improved, researchers face challenges in real-world applications in specific domains such as disaster simulation, healthcare, and urban planning due to data scarcity and unstable learning causing Image distortion. This paper proposes a new progressive learning methodology called Progressive Step Training (PST) based on the Auxiliary Classifier GAN (ACGAN) that discriminates class labels, leveraging the progressive learning approach of the Progressive Growing of GAN (PGGAN). The PST model achieves 70.82% faster stabilization, 51.3% lower standard deviation, stable convergence of loss values in the later high resolution stages, and a 94.6% faster loss reduction compared to conventional methods.
참고문헌 (28)
- Y. Liao, "Deep Learning-Based Application of Image Style Transfer", Mathematical Problems in Engineering, 2022(1693892), pp.1~2 (2022)[DOI]
- F. Ahmed, "Deep Learning Modelling Techniques : Current Progress, Applications, Advantages, and Challenges", Artificial Intelligence Review, 56, pp.13521~13617 (2023)[DOI]
- I. J. Goodfellow, "Generative Adversarial Networks", Advances in Neural Information Processing Systems, 27 (2014)
- G. C. Marano, "Generative Adversarial Networks Review in Earthquake-Related Engineering Fields", Bulletin of Earthquake Engineering, pp.1~52 (2023)[DOI]
- A. You, "Application of Generative Adversarial Networks(GAN)for Ophthalmology Image Domains : A Survey", Eye and Vision, 9(1), pp.1~19 (2022)
- A. Waheed, "CovidGAN : Data Augmentation Using Auxiliary Classifier GAN for Improved COVID-19 Detection", IEEE Access, 8, pp.91916~91923 (2020)[DOI]
- S. Fedorova, "Generative Adversarial Networks for Urban Block Design", SimAUD 2021: A Symposium on Simulation for Architecture and Urban Design, pp.1~10 (2021)
- J. Deng, "ImageNet:A Large-Scale Hierarchical Image Database", 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp.248~255 (2009), IEEE
- Y. Shoham, "Multiagent Systems : Algorithmic, Game-Theoretic, and Logical Foundations" (2008), Cambridge University Press
- A. Jabbar, "A Survey on Generative Adversarial Networks : Variants, Applications, and Training", ACM Computing Surveys, 54(8), pp.1~49 (2021)[DOI]
- Priyanka Rana, "Data augmentation with improved regularisation and sampling for imbalanced blood cell image classification", Scientific Reports, 12(1), pp.18101~ (2022)[DOI]
- Nagwa Elaraby, "A conditional GAN-based approach for enhancing transfer learning performance in few-shot HCR tasks", Scientific Reports, 12(1), pp.16271~ (2022)[DOI]
- Skandarani, "GANs for Medical Image Synthesis : An Empirical Study", Journal of Imaging, 9(3), pp.69~ (2023)[DOI]
- Y. Chen, "Dynamic Feature Splicing for Few-Shot Rare Disease Diagnosis", Medical Image Analysis, 90, pp.102959~ (2023)[DOI]
- Nishtha Jatana, "Future Frame Prediction using Generative Adversarial Networks", Karbala International Journal of Modern Science, 10(1), pp.2~ (2023)[DOI]
- Andreas Triantafyllopoulos, "An overview of affective speech synthesis and conversion in the deep learning era", Proceedings of the IEEE, 111(10), pp.1355~1381 (2023)[DOI]
- X. Rui, "DisasterGAN : Generative Adversarial Networks for Remote Sensing Disaster Image Generation", Remote Sensing, 13(21), pp.4284~ (2021)[DOI]
- A. Odena, "Conditional Image Synthesis with Auxiliary Classifier GANs", International Conference on Machine Learning, pp.2642~2651 (2017), PMLR
- H. Yang, "SPE-ACGAN : A Resampling Approach for Class Imbalance Problem in Network Intrusion Detection Systems", Electronics, 12(15), pp.3323~ (2023)[DOI]
- C. Han, "Neural Approaches to Dynamics of Signal Exchanges", pp.291~303 (2020)
- "Learning Multiple Layers of Features from Tiny Images"[URL]
- J. Gui, "A Review on Generative Adversarial Networks : Algorithms, Theory, and Applications", IEEE Transactions on Knowledge and Data Engineering, 35(4), pp.3313~3332 (2021)[DOI]
- M. Mohebbi Moghaddam, "Games of GANs : Game-theoretical Models for Generative Adversarial Networks", Artificial Intelligence Review, 56, pp.9771~9807 (2023)
- K. Kokomoto, "Intraoral Image Generation by Progressive Growing of Generative Adversarial Network and Evaluation of Generated Image Quality by Dentists", Scientific Reports, 11(1), pp.18517~ (2021)[DOI]
- T. Salimans, "Improved Techniques for Training GANs", Advances in Neural Information Processing Systems, 29 (2016)