An improved categorical cross entropy for remote sensing image classification based on noisy labels. (1st November 2022)

• Record Type: Journal Article
• Authors: Li, Panle; He, Xiaohui; Cheng, Xijie; Qiao, Mengjia; Song, Dingjun; Chen, Mingyang; Zhou, Tao; Li, Jiamian; Guo, Xiaoyu; Hu, Shaokai; Tian, Zhihui
• Is Part Of: Expert systems with applications. Volume 205(2022)
• Publication Date: 2022-11-01
• Subjects: Volunteered geographic information -- Remote sensing image classification -- Deep convolutional neural networks -- Noisy label -- Improved categorical cross-entropy -- Sample weighting scheme; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.117296
• Languages: English
• ISSNs: 0957-4174

Abstract

Abstract: In recent years, volunteered geographic information (VGI) is widely used to train deep convolutional neural networks (DCNNs) for high-resolution remote sensing image classification. However, noisy labels are often included in the training samples generated by VGI, which inevitably affects the performance of DCNNs. To solve this problem, the negative effect of noisy labels on remote sensing image classification by DCNNs is analyzed. Then, an improved categorical cross-entropy (ICCE) is proposed to address the issue of noisy labels. The ICCE improves the robustness of DCNN to noisy labels by revisiting the sample weighting scheme so that much attention is paid to the clean samples instead of the noisy samples. Besides, the error bound of ICCE is derived with strict mathematical proof under different types of noisy labels, which ensures the advantages of ICCE from theory. Additionally, extensive experiments are conducted on three remote sensing image datasets with simulated and real noisy labels to quantitatively evaluate the performance of ICCE. Highlights: This paper addresses the noisy label issue in remote sensing images classification. An improved categorical cross entropy (ICCE) is proposed to handle noisy labels. The theoretical error bound of ICCE is given with strict mathematical proof. We demonstrate the effectiveness of ICCE on three remote sensing image datasets.