대한임베디드공학회논문지 (IEMEK Journal of Embedded Systems and Applications)

대한임베디드공학회 (Institute of Embedded Engineering of Korea)
권호 표지
DOI QR코드

DOI QR Code

이미지 분류 성능 향상을 위한 무작위성 적용 방법론

Methodology of Applying Randomness for Boosting Image Classification Performance

  • 박주용 ;
  • 전유리 ;
  • 김미영 ;
  • 이정민 ;
  • 현윤석
  • Juyong Park (Inha University) ;
  • Yuri Jeon (Inha University) ;
  • Miyeong Kim (Inha University) ;
  • Jeongmin Lee (Inha University) ;
  • Yoonsuk Hyun (Inha University)
  • 투고 : 2024.06.28
  • 심사 : 2024.09.07
  • 발행 : 2024.10.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

Securing various types of training data in image Classification is important for improving performance. However, increasing the amount of original data is cost-limited, so data diversity can be secured by transforming images through data augmentation. Recently, a new deep learning approach using Generative AI models like GAN or Diffusion Based models has emerged in the Data Augmentation task, and reinforcement learning-based methods such as AutoAugment and Deep AutoAugment using existing basic Augmentation techniques are also showing good performance. However, this method has the disadvantage of having a complicated optimization procedure and high computational cost. This paper conducted various experiments with existing methods Cutmix, Mixup, RandAugment. By combining these techniques appropriately, we obtained higher performance than existing method without much effort. Additionally, our ablation experiment shows that additional hyper-parameter adjustments are needed for the basic augmentation types when we use RandAugment. Our code is available at https://github.com/lliee1/Randomness_Analysis.

키워드

과제정보

본 논문은 2022년, 2024년 정부 (과학기술정보통신부)의 재원으로 한국연구재단의 지원을 받아 수행된 연구임 (No. 2022R1A4A5033271, No.RS-2024-00348476).

참고문헌

  1. A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale," International Conference on Learning Representations, 2021.
  2. J. N. Chen, S. Sun, J. He, P. H. Torr, A. Yuille, S. Bai, "Transmix: Attend to Mix for Vision Transformers," Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 12135-12144, 2022.
  3. Q. Zhao, Y. Huang, W. Hu, F. Zhang, J. Liu, "Mixpro: Data Augmentation with Maskmix and Progressive Attention Labeling for Vision Transformer," arXiv preprint arXiv:2304.12043, 2023.
  4. H. Lee, H. Jung, "Recyclable Objects Detection Via Bounding Box CutMix and Standardized Distance-based IoU," IEMEK J. Embed. Sys. Appl., Vol. 17, No. 5, pp. 289-296, 2022.
  5. A. Bochkovskiy, C. Wang, H. M. Liao, "Yolov4: Optimal Speed and Accuracy of Object Detection," arXiv preprint arXiv:2004.10934, 2020.
  6. I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, Y. Bengio, "Generative Adversarial Nets," Advances in Neural Information Processing Systems, 27, 2014.
  7. R. Rombach, A. Blattmann, D. Lorenz, P. Esser, B. Ommer, "High-resolution Image Synthesis with Latent Diffusion Models," Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10684-10695. 2022.
  8. Y. Zhou, H. Sahak, J. Ba, "Training on Thin Air: Improve Image Classification with Generated Data," arXiv preprint arXiv:2305.15316, 2023.
  9. S. Motamed, P. Rogalla, F. Khalvati, "Data Augmentation Using Generative Adversarial Networks (GANs) For GAN-Based Detection Of Pneumonia And COVID-19 In Chest X-Ray Images," Informatics in Medicine Unlocked, 27, 100779, 2021.
  10. B. Trabucco, K. Doherty, M. Gurinas, R. Salakhutdinov, "Effective Data Augmentation with Diffusion Models," arXiv preprint arXiv:2302.07944, 2023.
  11. E. D. Cubuk, B. Zoph, D. Mane, V. Vasudevan, Q. V. Le, "Autoaugment: Learning Augmentation Policies from Data," arXiv preprint arXiv:1805.09501, 2018.
  12. Y. Zheng, Z. Zhang, S. Yan, M. Zhang, "Deep Autoaugment," arXiv preprint arXiv:2203.06172, 2022.
  13. S. Yun, D. Han, S. J. Oh, S. Chun, J. Choe, Y. Yoo, "Cutmix: Regularization Strategy to Train Strong Classifiers with Localizable Features," Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 6023-6032, 2019.
  14. H. Zhang, M. Cisse, Y. N. Dauphin, D. Lopez-Paz, "Mixup: Beyond Empirical Risk Minimization," arXiv preprint arXiv:1710.09412, 2017.
  15. E. D. Cubuk, B. Zoph, J. Shlens, Q. V. Le, "Randaugment: Practical Automated Data Augmentation with a Reduced Search Space," Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pp. 702-703, 2020.
  16. K. He, X. Zhang, S. Ren, J. Sun, "Deep Residual Learning for Image Recognition," Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770-778, 2016.
  17. S. Zagoruyko, N. Komodakis, "Wide Residual Networks," arXiv preprint arXiv:1605.07146, 2016.
  18. Y. Le, X. Yang, "Tiny Imagenet Visual Recognition Challenge," CS 231N, 7.7: 3, 2015.
  19. A. Krizhevsky, G. Hinton, "Learning Multiple Layers of Features from Tiny Images," 2009.
  20. T. DeVries, G. W. Taylor, "Improved Regularization of Convolutional Neural Networks with Cutout," arXiv preprint arXiv:1708.04552, 2017.
  21. V. Hosu, H. Lin, T. Sziranyi, D. Saupe, "KonIQ-10k: An Ecologically Valid Database for Deep Learning of Blind Image Quality Assessment," IEEE Transactions on Image Processing, Vol. 29, pp. 4041-4056, 2020.
  22. B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, A. Torralba, "Learning Deep Features for Discriminative Localization," Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2921-2929, 2016.