Fig. 1. Example of a DenseNet with four layers (i = 4)

Fig. 2. Subspace feature transformation

Fig. 3. Schematic of the hyperspectral image classification based on DCDA

Fig. 4. Schematic of the dense convolution-based embedding module (The number of convolutional layers is 3)

Fig. 5. Schematic of the discriminator module

Fig. 6. Source and target images in Indiana dataset

Fig. 7. Source and target images in Pavia dataset

Fig. 8. Source and target image spectral curves

Fig. 9. Classification result of target Indiana dataset

Fig. 10. Classification result of target Indiana dataset