임계값 학습 모듈을 적용한 준지도 SAR 이미지 분류
Semi-supervised SAR Image Classification with Threshold Learning Module
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
준지도 학습(Semi-supervised learning)은 소량의 라벨이 있는 데이터와 다량의 라벨이 없는 데이터를 사용하여모델을 훈련하는 효과적인 방법이다. 그러나 많은 논문에서 준지도 학습시 하나의 고정된 임계값을 사용하여각 클래스별 서로 다른 이미지들의 특징별 차이를 고려하지 않고 임의 라벨을 만든다. 본 논문에서는 합성개구레이더(SAR) 이미지 분류 준지도 학습시 모든 클래스가 하나의 고정된 임계값을 사용하는 대신 각 클래스에 대해 서로 다른 임계값을 적용한다. 모델에 임계값 학습 모듈을 추가하여 임계값을 학습하여 클래스별로 학습되는 차이를 고려하여 클래스별로 서로 다른 임계값을 얻는다. 서로 다른 임계값을 사용한 준지도 학습 기반의 SAR 이미지 분류 방법을 적용유무를 비교하여 클래스별 임계값을 사용하는 이점에 대해 고찰하였다.
Abstract
Semi-supervised learning (SSL) is an effective approach to training models using a small amount of labeled data and a larger amount of unlabeled data. However, many papers in the field use a fixed threshold when applying pseudo-labels without considering the feature-wise differences among images of different classes. In this paper, we propose a SSL method for synthetic aperture radar (SAR) image classification that applies different thresholds for each class instead of using a single fixed threshold for all classes. We propose a threshold learning module into the model, considering the differences in feature distributions among classes, to dynamically learn thresholds for each class. We compare the application of a SSL SAR image classification method using different thresholds and examined the advantages of employing class-specific thresholds.
참고문헌 (27)
- F. Pierre, "Sharpness-aware minimization for efficiently improving generalization", arXiv preprint arXiv:2010.01412, ~ (2020)
- M. Wortsman, "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time", International Conference on Machine Learning (2022), PMLR
- H. M. Kabir, "Reduction of Class Activation Uncertainty with Background Information", arXiv preprint arXiv:2305.03238, ~ (2023)
- Z, Yang, "SAR image classification method based on improved capsule network", Journal of Physics: Conference Series, 1693(1) (2020), IOP Publishing
- S, Chen, "Target classification using the deep convolutional networks for SAR images", IEEE transactions on geoscience and remote sensing, 54(8), pp.4806~4817 (2016)
- H, Ren, "Extended convolutional capsule network with application on SAR automatic target recognition", Signal Processing, 183, pp.108021~ (2021)
- Q, Xie, "Self-training with noisy student improves imagenet classification", Proceedings of the IEEE, 2020, ~
- E, Arazo, "Pseudo-labeling and confirmation bias in deep semi-supervised learning", 2020 International Joint Conference on Neural Networks (IJCNN) (2020), IEEE
- D, Berthelot, "Mixmatch: A holistic approach to semi-supervised learning", Advances in neural information processing systems, 32 (2019)
- D, Berthelot, "Remixmatch : Semi-supervised learning with distribution alignment and augmentation anchoring", arXiv preprint arXiv:1911.09785, ~ (2019)
- K. Sohn, Kihyuk, "Fixmatch: Simplifying semi-supervised learning with consistency and confidence", Advances in neural information processing systems, 33, pp.596~608 (2020)
- T. Chen, "A s imple framework for contrastive learning of visual representations", International conference on machine learning (2020), PMLR
- A. Anaby-Tavor, "Do not have enough data? Deep learning to the rescue!", Proceedings of the AAAI Conference on Artificial Intelligence, 34(05) (2020)
- B. Zhang, "Censer: Curriculum semi-supervised learning for speech recognition based on self-supervised pre-training", arXiv preprint arXiv:2206.08189, ~ (2022)
- V. Tsouvalas, "Federated self-training for semi-supervised audio recognition", ACM Transactions on Embedded Computing Systems, 21(6), pp.1~26 (2022)
- M. Sajjadi, "Regularization with stochastic transformations and perturbations for deep semi-supervised learning", Advances in neural information processing systems, 29 (2016)
- X. Wang, "Contrastive learning with stronger augmentations", IEEE Transactions on Pattern Analysis and Machine Intelligence (2022)
- E. Cubuk, "Randaugment: Practical automated data augmentation with a reduced search space", Proceedings of the IEEE, 2020, ~
- E.D. Cubuk, "Autoaugment: Learning augmentation strategies from data", Proceedings of the IEEE, 2019, ~
- Y. Lei, "Synthetic Images Augmentation for Robust SAR Target Recognition", 2021 The 5th International Conference on Video and Image Processing (2021)
- X. Zhang, "A Novel Data Augmentation Method for SAR Image Target Detection and Recognition", 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS (2021), IEEE
- M. Zhang, "Data augmentation method of SAR image dataset", IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium (2018), IEEE
- Xu, Yi, "Dash: Semi-supervised learning with dynamic thresholding", International Conference on Machine Learning (2021), PMLR
- Liang, Zechen, "ADT-SSL: Adaptive Dual-Threshold for Semi-Supervised Learning", arXiv preprint arXiv:2205.10571, ~ (2022)
- Zhang, Bowen, "Flexmatch: Boosting semi-supervised learning with curriculum pseudo labeling", Advances in Neural Information Processing Systems, 34, pp.18408~18419 (2021)
- E.R. Keydel, "MSTAR extended operating conditions: A tutorial", Algorithms for Synthetic Aperture Radar Imagery III, 2757, pp.228~242 (1996)
- C. Coman, "A deep learning SAR target classification experiment on MSTAR dataset", 2018 19th international radar symposium (IRS) (2018), IEEE