Figure 5
From: A new strategy to map landslides with a generalized convolutional neural network
Result of applying the trained CNN model (\(M_{6}^{ALL}\)) on study area “E” to “G”. These three study areas were not used for training any CNN model; hence they were completely used as the testing region. The results highlight the generalizing ability of \(M_{6}^{ALL}\) to map event landslides. Every row shows a fixed view from one study area with pre-event image (left), post-event image with polygons from landslide inventory (middle), and output from CNN as a confusion matrix map (right). The majority resampling method was used to visualize the confusion matrix map. The cloud and snow-cover masks have not been shown for clarity, but the affected region was not used in the training and prediction process. The maps were created using ArcGIS Pro v2.4 (https://www.esri.com/).
