Fig. 1: Schematic of the procedures to quantify delta morphologies.
From: Identifying controlling factors of delta morphology using a convolutional autoencoder
The network architecture of the convolutional autoencoder (CAE) model was constructed to output images similar to the input images, which indicates that the model compressed the input image into a 70-dimensional latent code at the middle of the network. Examples of the original and predicted delta images are shown on the left and right sides of the CAE model, respectively. The darker color in the predicted images indicates a higher probability that the pixels are classified into land regions.
