Figure 10
From: Synthetic aircraft RS image modelling based on improved conditional GAN joint embedding network
Embedding results and loss value on MNIST and Fashion-MNIST datasets. (a) The red frames represent the ground-truth 32 pixels \(\times\) 32 pixels images x, and columns 2 to 10 represent the embedding results, \(G\left( z^{*}, y\right)\), by the GE with various class labels y. (b) Embedding loss values vs. the number of optimisation steps on MNIST. (c) Embedding loss values vs. the number of optimisation steps on Fashion-MNIST datasets.
