Fig. 1. Schematic of single-pixel imaging

Fig. 2. Driving styles in deep learning single pixel imaging. (a) Data-driven; (b) physically-driven

Fig. 3. Deep learning of the role of neural networks in single pixel imaging. (a) "Image to image" method; (b) "measurements to image" method

Fig. 4. Neural networks for"image-to-image"style data-driven deep learning in single pixel imaging^[26]. (a) Fully connected layer neural network; (b) convolutional neural network

Fig. 5. Single pixel imaging reconstruction results of neural networks for"image-to-image"data-driven deep learning^[27]

Fig. 6. Schematic of the end-to-end deep learning single-pixel imaging^[46]

Fig. 7. Schematic of autoencoder learning sampling module and reconstruction module^[56]

Fig. 8. Photoelectric diagram of data-driven deep learning single-pixel target detection^[63]. (a) Network training; (b) proposed imaging-free method; (c) conventional image-based method

Fig. 9. Schematic of physically constrained neural network single-pixel imaging method^[80]

Fig. 10. Schematic of "measurements to image" physical constrained neural network single-pixel imaging method^[84]

Fig. 11. Schematic of neural network single pixel imaging method fitting forward physical model^[88]

Fig. 12. Schematic of data prior physical enhancement single-pixel imaging method^[89]. (a) Training process of data-driven network; (b) single pixel imaging system; (c) physical model-drive fine tuning

Fig. 13. Schematic of the single-pixel imaging method of prior physical enhancement generated by large model^[92]