Multiview Deep Learning for Land-Use Classification

Abstract

Abstract-A multiscale input strategy for multiview deep learning is proposed for supervised multispectral land-use classification and it is validated on a well-known dataset. The hypothesis that simultaneous multiscale views can improve compositionbased inference of classes containing size-varying objects compared to single-scale multiview is investigated. The end-to-end learning system learns a hierarchical feature representation with the aid of convolutional layers to shift the burden of feature determination from hand-engineering to a deep convolutional neural network. This allows the classifier to obtain problemspecific features that are optimal for minimizing the multinomial logistic regression objective, as opposed to user-defined features which trades optimality for generality. A heuristic approach to the optimization of the deep convolutional neural network hyperparameters is used, based on empirical performance evidence. It is shown that a single deep convolutional neural network can be trained simultaneously with multiscale views to improve prediction accuracy over multiple single-scale views. Competitive performance is achieved for the UC Merced dataset where the 93.48% accuracy of multiview deep learning outperforms the 85.37% accuracy of SIFT-based methods and the 90.26% accuracy of unsupervised feature learning.Index Terms-Neural network applications, neural network architecture, feature extraction, urban areas, remote sensing.

Deep Learning in Remote Sensing: A Comprehensive Review and List of Resources

“…The deep arhitectures discussed in Section II-B have been applied to the problem of scene classification of high-resolution satellite images and led to state-of-the-art performance [71,74,[80][81][82][83][84][85][86][87]. As deep learning is a multi-layer feature learning architecture, it can learn more abstract and discriminative semantic features as the depth grows and achieve far better classification performance compared with the mid-level approaches.…”

Pre-Trained AlexNet Architecture with Pyramid Pooling and Supervision for High Spatial Resolution Remote Sensing Image Scene Classification

“…Pre-trained AlexNet 1st-FC [32] 95.08 Multiview deep learning [31] 93. 48 From Table 1, it can be seen that the pre-trained-AlexNet-SPP and the pre-trained-AlexNet-SS obtain better scene classification performances than the pre-trained-AlexNet architecture, which proves that the incorporation of either SPP or SS can improve the scene classification performance.…”

Scene Classification Based on a Deep Random-Scale Stretched Convolutional Neural Network

“…2017, 9, x FOR PEER REVIEW 10 of 23 (SSBFC) [7], locality-constrained linear coding (LLC) [7], SPM+SIFT [7], SAL-PTM [10], the Dirichletderived multiple topic model (DMTM) [14], , SIFT+SC [55], the local Fisher kernel-linear (LFK-Linear), the Fisher kernel-linear (FK-Linear), the Fisher kernel incorporating spatial information (FK-S) [56], and methods based on deep learning, i.e., saliency-guided unsupervised feature learning (S-UFL) [37], the gradient boosting random convolutional network (GBRCN) [40], the large patch convolutional neural network (LPCNN) [41], and the multiview deep convolutional neural network (M-DCNN) [36], were compared. In the experiments, 80% of the samples were randomly selected from the dataset as the training samples, and the rest were used as the test samples.…”

“…Deep learning methods M-DCNN [36] 93.48 S-UFL [37] 82.72 GBRCN [40] 94.53 LPCNN [41] 89 From Figure 7a, it can be seen that the 21 classes can be recognized with an accuracy of at least 85%, and most categories are recognized with an accuracy of 100%, i.e., agriculture, airplane, baseball diamond, beach, etc. From Figure 7b, it can be seen that some representative scene images on the left are recognized correctly by SRSCNN and CNNV.…”