Fig. 1. Diagram of structures, principle, and imaging result of common optical fibers. (a) Single-mode fiber; (b) multi-core fiber; (c) multi-mode fiber

Fig. 2. Typical deep learning models in intelligent fiber optic imaging. (a) Fully connected neural network^[64]; (b) U-Net^[65]; (c) GAN^[66]

Fig. 3. Incoherent imaging based on multi-core optical fiber^[79]. (a) In-line dual-sensor fiber imaging system; (b) overall structure of the image restoration model based on MBIN (model-based iterative network); (c) incoherent imaging results based on multi-core optical fiber and reconstruction images by deep learning after splicing

Fig. 4. Intelligent phase imaging system based on multi-core fiber^[4]. (a) Data acquisition system of quantitative phase imaging based on multi-core fiber; (b) traditional iterative quantitative phase imaging algorithm procedure; (c) quantitative phase imaging reconstruction model based on U-Net

Fig. 5. Disordered Anderson positioning optical fiber imaging system driven by deep learning^[87]. (a) Imaging system based on disordered Anderson localizing fiber; (b) disordered Anderson localizing fiber speckle reconstruction model based on deep learning; (c) real images of various cells, corresponding disordered Anderson fiber speckle images and reconstructed images

Fig. 6. Intelligent speckle imaging system based on multimode fiber^[60,98-99]. (a) Multi-mode fiber speckle imaging system; (b) speckle reconstruction model based on VGG; (c) speckle reconstruction model based on U-Net; (d) speckle reconstruction model based on GAN

Fig. 7. Speckle reconstruction diagram of handwritten Latin letters based on multi-mode fiber^[98]. (a) Model reconstructs input amplitude image from output amplitude speckle; (b) model reconstructs input phase image from output amplitude speckle; (c) model predicts corresponding spatial light modulator phase image through real image, and real image is reconstructed on the camera after passing through the multi-mode fiber

Fig. 8. Image reconstruction system of one-dimensional temporal signal based on multi-mode optical fiber^[110]. (a) Data acquisition system of one-dimensional temporal signal corresponding to real image, and image reconstruction model; (b) real image, its corresponding one-dimensional temporal signal and reconstructed image

Fig. 9. Application of multi-mode fiber imaging in computer vision tasks^[112-114]. (a) Image classification based on multi-mode fiber speckle; (b) edge detection based on multi-mode fiber speckle; (c) image denoising based on multi-mode fiber speckle

Fig. 10. Comparison of cancer diagnosis biopsy workflows^[83]. (a) Traditional biopsy procedure; (b) end-to-end biopsy based on intelligent lensless fiber imaging