Attention Consistent Network for Remote Sensing Scene Classification

Abstract

Remote sensing (RS) image scene classification is an important research topic in the RS community, which aims to assign the semantics to the land covers. Recently, due to the strong behavior of convolutional neural network (CNN) in feature representation, the growing number of CNN-based classification methods has been proposed for RS images. Although they achieve cracking performance, there is still some room for improvement. First, apart from the global information, the local features are crucial to distinguish the RS images. The existing networks are good at capturing the global features since the CNNs' hierarchical structure and the nonlinear fitting capacity. However, the local features are not always emphasized. Second, to obtain satisfactory classification results, the distances of RS images from the same/different classes should be minimized/maximized. Nevertheless, these key points in pattern classification do not get the attention they deserve. To overcome the limitation mentioned above, we propose a new CNN named attention consistent network (ACNet) based on the Siamese network in this article. First, due to the dual-branch structure of ACNet, the input data are the image pairs that are obtained by the spatial rotation. This helps our model to fully explore the global features from RS images. Second, we introduce different attention techniques to mine the objects' information from RS images comprehensively. Third, considering the influence of the spatial rotation and the similarities between RS images, we develop an attention consistent model to unify the salient regions and impact/separate the RS images from the same/different semantic categories. Finally, the classification results can be obtained using the learned features. Three popular RS scene datasets are selected to validate our ACNet.

First and Second-Order Information Fusion Networks for Remote Sensing Scene Classification

“…The skip-connected covariance (SCCov) network [12] directly uses covariance matrix of different convolution features as the image representation. Other works include PANet50 [19], ResNet-101+EAM [20] and ACNet [21], all focusing on the selfattention-based fusion strategies to enhance feature representations, and have achieved competitive performance. Amongst these methods, SCCov, as a typical second-order pooling method, cannot achieve better performances than other methods.…”

A Lightweight Model of VGG-16 for Remote Sensing Image Classification

“…At the same time, we input the data set selected by our model into the newly published (Attention Consistent Network for Remote Sensing Scene Classification) [38] The models and algorithms in classification are mainly for high-precision image classification, and because the pixels of the model we used are too low and the similarity is very high, there is over fitting phenomenon when the model runs EuroSAT dataset. [42]and LEVIR (800 × 600) [43] are selected to test the model.…”

Hierarchical Self-Attention Embedded Neural Network With Dense Connection for Remote-Sensing Image Semantic Segmentation

“…Attention is widely used for various tasks, such as machine translation [52], scene classification [53], and semantic segmentation [54]. The early attention mechanism was only designed to learn channel-wise correlations.…”