Figure 6 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Classification maps of different CNN methods for the training area: (a) 2D CNN, (b) 3D CNN, and (c) HybridSN.
Source publication
In recent years, deep learning-based image classification has become widespread, especially in remote sensing applications, due to its automatic and strong feature extraction capability. However, as deep learning methods operate on rectangular-shaped image patches, they cannot accurately extract objects’ boundaries, especially in complex urban sett...
Context in source publication
Context 1
... proposed classification workflow experimented with three CNN models, namely, 2D CNN, 3D CNN, and HybridSN. A visual comparison of these methods as classification maps is presented in Figures 6 and 7 for the training and test areas, respectively. Table 4 presents the classification accuracies obtained for the three methods in the training and test areas including the overall accuracies (OA, Kappa, mIoU) as well as perclass accuracies. ...