Fig. 1. The overall working principle of the GAN-guided DNN. (a) GAN-guided DNN for visual tracking and imaging of the interested moving target. (b) The training process of the GAN-guided DNN.

Fig. 2. The process of training the GAN-based teacher model, which involves acquiring datasets and optimizing models. (a) The principle of input dataset acquisition using the event-based camera. (b) The architecture of the GAN-based teacher model.

Fig. 3. Simulation results of GAN-guided DNN. (a) Examples of training results for the visual tracking and imaging of the target car. (b) The phase profiles of diffractive layers after deep learning-based optimization. (c) The PSNR and SSIM values with different input images.

Fig. 4. The GAN-guided DNN trained and tested with different numbers of diffractive layers. (a) The performance of the GAN-guided DNN with different numbers of diffractive layers ($L$) for the scenario with just a pedestrian and some static objects. (b) The SSIM and PSNR values with different input images.

Fig. 5. Experimental demonstration of the visual tracking using a GAN-guided DNN. (a) Schematic diagram of the experimental setup and phase used in the experiment (layer 1 and layer 2 are loaded on SLM 1 and SLM 2, respectively). HWP, half-wave plate; PBS, polarization beam splitter; QWP, quarter-wave plate; BS, beam splitter; SLM, spatial light modulator. (b) Simulation and experimental results of visual tracking and imaging of the target airplane in a scenario involving airplanes and missiles. (c) The SSIM and PSNR values of the simulation and experimental results with different input images.