Abstract: In the realm of scene classification, it is clear that deep learning models excel when a large amount of labeled data is available. However, continual learningsuffers from a lack of old tasks' data. Continual learning (CL) is a needed aspect of artificial intelligence (AI). Inspired from the human ongoing learning capacity, endowing a deep learning model with the ability to preserve previous knowledge is legitimate. Training a deep learning model for sequential learning of tasks leads to a continual decline in the performance for previous tasks due to the non-availability of their training data. This phenomenon is known, in the literature, as catastrophic forgetting. We proposed a two deep blocks.......
Keywords: Remote sensing, scene classification, data regeneration, contrastive learning, continual Learning.
[1] L. Zhao, P. Tang, and L. Huo, "Feature significance-based multibag-of-visual-words model for remote sensing image scene classification," Journal of Applied Remote Sensing, vol. 10, no. 3, p. 035004, 2016.
[2] J. Shi, Z. Jiang, H. Feng, and Y. Ma, "Sparse coding-based topic model for remote sensing image segmentation," presented at the 2013 IEEE International Geoscience and Remote Sensing Symposium-IGARSS, IEEE, 2013, pp. 4122–4125.
[3] G. Cheng, J. Han, L. Guo, Z. Liu, S. Bu, and J. Ren, "Effective and efficient midlevel visual elements-oriented land-use classification using VHR remote sensing images," IEEE Transactions on Geoscience and Remote Sensing, vol. 53, no. 8, pp. 4238–4249, 2015.
[4] J. Määttä, A. Hadid, and M. Pietikäinen, "Face spoofing detection from single images using micro-texture analysis," presented at the 2011 international joint conference on Biometrics (IJCB), IEEE, 2011, pp. 1–7.
[5] Q. Wang, W. Huang, Z. Xiong, and X. Li, "Looking Closer at the Scene: Multiscale Representation Learning for Remote Sensing Image Scene Classification," IEEE Transactions on Neural Networks and Learning Systems, 2020.
