The multimode fiber (MMF) has great potential to transmit high-resolution images with less invasive methods in endoscopy due to its large number of spatial modes and small core diameter. However, spatial modes crosstalk will inevitably occur in MMFs, which makes the received images become speckles. A conditional generative adversarial network (GAN) composed of a generator and a discriminator was utilized to reconstruct the received speckles. We conduct an MMF imaging experimental system of transmitting over 1 m MMF with a 50 μm core. Compared with the conventional method of U-net, this conditional GAN could reconstruct images with fewer training datasets to achieve the same performance and shows higher feature extraction capability.

Photonic waveguide arrays provide a simple and versatile platform for simulating conventional topological systems. Here, we investigate a novel one-dimensional (1D) topological band structure, a dimer chain, consisting of silicon waveguides with alternating self-coupling and inter-coupling. Coupled mode theory is used to study topological features of such a model. It is found that topological invariants of our proposed model are described by the global Berry phase instead of the Berry phase of the upper or lower energy band, which is commonly used in the 1D topological models such as the Su–Schrieffer–Heeger model. Next, we design an array configuration composed of two dimer patterns with different global Berry phases to realize the topologically protected waveguiding. The topologically protected propagation feature is simulated based on the finite-difference time-domain method and then observed in the experiment. Our results provide an in-depth understanding of the dynamics of the topological defect state in a 1D silicon waveguide array, and may provide different routes for on-chip lightwave shaping and routing.

We propose here a novel method for position fixing in the micron scale by combining the convolutional neural network (CNN) architecture and speckle patterns generated in a multimode fiber. By varying the splice offset between a single mode fiber and a multimode fiber, speckles with different patterns can be generated at the output of the multimode fiber. The CNN is utilized to learn these specklegrams and then predict the offset coordinate. Simulation results show that predicted positions with the precision of 2 μm account for 98.55%. This work provides a potential high-precision two-dimensional positioning method.

Inspired by recent rapid deep learning development, we present a convolutional-neural-network (CNN)-based algorithm to predict orbital angular momentum (OAM) mode purity in optical fibers using far-field patterns. It is found that this image-processing-based technique has an excellent ability in predicting the OAM mode purity, potentially eliminating the need of using bulk optic devices to project light into different polarization states in traditional methods. The excellent performance of our algorithm can be characterized by a prediction accuracy of 99.8% and correlation coefficient of 0.99994. Furthermore, the robustness of this technique against different sizes of testing sets and different phases between different fiber modes is also verified. Hence, such a technique has a great potential in simplifying the measuring process of OAM purity.

A novel see-through virtual retina display (VRD) system is proposed in this Letter. An optical fiber projector is used as the thin-light-beam source, which is modified from a laser scan projector by separating the laser sources and the scan mechanical structure. A synthetic aperture method is proposed for simple, low-cost fabrication of a volume holographic lens with large numerical aperture. These two key performance-enhanced elements are integrated into a lightweight and ordinary-glasses-like optical see-through VRD system. The proposed VRD system achieves a weight of 30 g and a diagonal field of view of 60°.

We propose an adaptive parallel coordinate (APC) algorithm for quickly forming a series of focused spots at a multimode fiber (MMF) output by controlling the MMF input field with a spatial light modulator (SLM). Only passing over the SLM once, we can obtain SLM reflectance to form focused spots on different positions. Compared with the transmission matrix method, our APC does not require iterations and massive calculations. The APC does not require as much access device time as the adaptive sequential coordinate ascent (SCA) algorithm. The experiment results demonstrate that the time taken to form 100 spots with our APC is 1/54th the time with the SCA.