Fig. 1. The 71 spectra classified refering to Tab.1

Fig. 2. Visualization diagrams of dimensionality reduction features obtained from real samples using various methods

Fig. 3. Silhouette coefficient of the real spectra

Fig. 4. Visualization of clustering results with K-means (a) and GMM (b) of Real samples

Fig. 5. Two spectral graphs with different clustering results from K-means and GMM

Fig. 6. Samples of spectra generated by DCGAN trained by all real spectra

Fig. 7. Visualization of the 120 spectra generated by DCGAN trained by all real spectra after dimensionality reduction

Fig. 8. Visualization of 120 generated spectra and 71 real ones after dimensionality reduction(The triangles are the generated spectra)

Fig. 9. A comparison of the spectral graph (a) generated by DCGAN after 50 iterations, the spectral graph (b) generated after 400 iterations, and the real spectral graph (c)

Fig. 10. real spectral graphs are analyzed separately with 25 (a) and 5 (b) spectral graphs generated by DCGAN trained on single-class spectral graphs with VGG16-PCA-visualization, with triangles indicating the generated spectral graphs

Fig. 11. The real spectral graphs are analyzed separately with the spectral graphs generated by DCGAN trained on single-class spectral graphs after 50 iterations (a) and 400 iterations (b) with VGG16-PCA-visualization. The triangles are the generated spectra