Yu Ma
ABSTRACT
To solve the problem that the data annotation in image classification task requires a lot of time and economic costs, and a large number of unlabeled images cannot be effectively utilized in reality, an image classification method combining active learning algorithm and data augmentation is proposed. First, data augmentation is performed on a small number of labeled samples, the classification model is initially trained, and then, according to the sampling strategy of active learning, the samples are selected and labeled by experts, which are the most conducive to model training from the unlabeled set. The labeled set is updated by adding the new labeled samples. The same process is performed until the requirements are met. In this paper, experiments are carried out on the digits dataset and the Cifar10 set. The results show that the image classification method proposed in this paper can effectively enhance the accuracy of image classification.
- Chifeng, C. and Dingkuo, H. 2012. Supervised learning method and quality capability of process used in an optical transmission inspection of on-line nonwoven basis weight. J. Optics Communications. 285, 8 (April. 2012) 2106--2112.Google Scholar
- Duong, N., Plessis, D., and Christoffel, M., et al. 2015. Continuous target shift adaptation in supervised learning. In 7th Asian Conference on Machine Learning (Hong Kong, November 20 - 22, 2015). Asian Conference on Machine Learning, 285--300.Google Scholar
- Hastie, T., Tibshirani, R., and Friedman, J. 2009. The elements of statistical learning: data mining, in-ference, and prediction. 2nd ed. Springer, New York.Google Scholar
- Kaiming, H., Xiangyu, Z., and Shaoqing, R., et al. 2016. Deep residual learning for image recognition. In Proceedings of the 29th IEEE Conference on Computer Vision and Pattern Recognition (Las Vegas, CA, United states, June 26 2016 - July 1, 2016). IEEE Computer Society, 770--778.Google Scholar
- Tuia, D., Volpi, M., and Copa, L., et al. 2011. A survey of active learning algorithms for supervised remote sensing image classification. J. IEEE Journal of selected Topics in Signal Processing. 5, 3 (June. 2011), 606--617.Google ScholarCross Ref
- Nguyen, L., Huong, M., and Albert, B., et al. 2019. Semi-supervised Learning over Streaming Data using MOA. In Proceedings of 2019 IEEE International Conference on Big Data (Los Angeles, CA, United states, December 9 - 12, 2019). Institute of Electrical and Electronics Engineers Inc, 553--562.Google ScholarCross Ref
- Inae, C., Kanghee, P., and Hyunjung, S. 2013. Sharpened graph ensemble for semi-supervised learning. J. Intelligent Data Analysis. 17, 3 (2013), 387--389.Google ScholarCross Ref
- Lei, Z., Yan, T., and Qiang, J. 2008. Active image labeling and its application to facial action labeling. In Proceedings of the 10th European Conference on Computer Vision (Marseille, France, October 12 - 18, 2008). Springer Verlag, 706--719.Google Scholar
- Ksenia, K., Raphael, S., and Pascal, F. 2017. Learning Active Learning from Data. In 31st Conference on Neural Information Processing Systems (Long Beach, CA, United states, December 4 - 9, 2017). Neural information processing systems foundation, 4228--4238.Google Scholar
- Ebert, S., Fritz, M., and Schiele, B. 2012. RALF: A reinforced active learning formulation for object class recognition. In Proceedings of the 25th IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Providence, RI, United states, June 16 - 21, 2012). IEEE Computer Society, 3626--3633.Google Scholar
- Dim P, P., Jasper R. R, U., and Frank, K., Vittorio F. 2016. We Don't need no bounding-boxes: Training object class detectors using only human verification. In Proceedings of the 29th IEEE Conference on Computer Vision and Pattern Recognition (Las Vegas, CA, United states, December 9, 2016). IEEE Computer Society, 854--863.Google Scholar
- Donggeun, Y. and In So, K. 2019. Learning loss for active learning. In Proceedings of the 32th IEEE Conference on Computer Vision and Pattern Recognition (Long Beach, CA, United states, June 16 - 20, 2019). IEEE Computer Society, 93--102.Google Scholar
- Wang, K., Yan, X., and Zhang, D., et al. 2018. Towards Human-Machine Cooperation: Self-Supervised Sample Mining for Object Detection. In Proceedings of the 31th IEEE Conference on Computer Vision and Pattern Recognition (Salt Lake City, UT, United states, June 18 - 22, 2018). IEEE Computer Society, 1605--1613.Google Scholar
- Chiehchi, K., Teegyok, L., and Pradeep, S., et al. 2018 Localization-Aware Active Learning for Object Detection. In: 14th Asian Conference on Computer Vision (Perth, WA, Australia). Springer Verlag, 506--522.Google Scholar
- Mosinska, A., Sznitman, R., and Glowacki, P., et al. 2016. Active Learning for delineation of curvilinear structures. In Proceedings of the 29th IEEE Conference on Computer Vision and Pattern Recognition (Las Vegas, CA, United states, June 26, 2016 - July 1, 2016). IEEE Computer Society, 5231--5239.Google Scholar
- Yi, Y., Zhigang, Z., and Feiping, N., et al. 2015. Multi-Class Active Learning by Uncertainty Sampling with Diversity Maximization. J. International Journal of Computer Vision. 113, 2 (June 16, 2015)113--127.Google Scholar
- Xin, L. and Yuhong, G. 2013. Adaptive active learning for image classification. In Proceedings of the 26th IEEE Conference on Computer Vision and Pattern Recognition (Portland, OR, United states, June 23 - 28, 2013). IEEE Computer Society, 859--866.Google Scholar
- Hongyi, Z., Moustapha, C., and Yann N, D., et al. 2018. MixUp: Beyond empirical risk minimization. In 6th International Conference on Learning Representations (Vancouver, BC, Canada, April 30, 2018 - May 3, 2018). International Conference on Learning Representations.Google Scholar
Index Terms
- Combining Active Learning and Data Augmentation for Image Classification
Recommendations
Multi-Label Active Learning Algorithms for Image Classification: Overview and Future Promise
Image classification is a key task in image understanding, and multi-label image classification has become a popular topic in recent years. However, the success of multi-label image classification is closely related to the way of constructing a training ...
Column subset selection for active learning in image classification
Image classification is an important task in computer vision and machine learning. However, it is known that manually labeling images is time-consuming and expensive, but the unlabeled images are easily available. Active learning is a mechanism which ...
Deep active learning models for imbalanced image classification
AbstractActive learning can query valuable samples in an unlabeled sample pool for annotation, thus building a more informative labeled dataset and reducing the annotation cost. However, traditional active learning methods are not effective in ...
Comments