Fig. 2 | Nature Communications

Fig. 2

From: Machine-learning reprogrammable metasurface imager

Fig. 2

Configurations and results for human body imaging. (a) and (b) report 16 leading modes of theoretical PCA and the corresponding experimental patterns radiated by the machine-learning imager, respectively. c shows the coding patterns of 2-bit coding metasurface corresponding to b. d shows four images of a testing person (Ya Shuang, coauthor), where the top two images are armed with glass scissors, as shown in red. e and f are the reconstructed images corresponding to d using the machine-learning imagers trained with PCA and random projection, respectively, where 400 measurements are used. Note that the image results have been normalized by their own maximums