Efficiently generating 3D holograms is one of the most challenging research topics in the field of holography. This work introduces a method for generating multi‐depth phase‐only holograms using a fully convolutional neural network (FCN). The method primarily involves a forward–backward‐diffraction framework to compute multi‐depth diffraction fields, along with a layer‐by‐layer replacement method (L2RM) to handle occlusion relationships. The diffraction fields computed by the former are fed into the carefully designed FCN, which leverages its powerful non‐linear fitting capability to generate multi‐depth holograms of 3D scenes. The latter can smooth the boundaries of different layers in scene reconstruction by complementing information of occluded objects, thus enhancing the reconstruction quality of holograms. The proposed method can generate a multi‐depth 3D hologram with a PSNR of 31.8 dB in just 90 ms for a resolution of 2160 × 3840 on the NVIDIA Tesla A100 40G tensor core GPU. Additionally, numerical and experimental results indicate that the generated holograms accurately reconstruct clear 3D scenes with correct occlusion relationships and provide excellent depth focusing.
A phase-only hologram generated through the convolution neutral network (CNN) which is trained by the low-frequency mixed noise (LFMN) is proposed. Compared with CNN based computer-generated holograms, the proposed training dataset named LFMN includes different kinds of noise images after low-frequency processing. This dataset was used to replace the real images used in the conventional hologram to train CNN in a simple and flexible approach. The results revealed that the proposed method could generate a hologram of 2160 × 3840 pixels at a speed of 0.094 s/frame on the DIV2K valid dataset, and the average peak signal-to-noise ratio of the reconstruction was approximately 29.2 dB. The results of optical experiments validated the theoretical prediction. The reconstructed images obtained using the proposed method exhibited higher quality than those obtained using the conventional methods. Furthermore, the proposed method considerably mitigated artifacts of the reconstructed images.
In this paper, the phase-only holograms with clear first diffraction orders have been generated based on the U-Net and the two-dimensional phase grating. Firstly, we proved the modulation effect of two-dimensional phase grating on diffraction field, and came to a conclusion that it could move the diffraction pattern of the hologram to the odd-numbered diffraction orders’ center of that. Then we changed the generation process of phase-only holograms and the training strategy for U-Net according to this conclusion, which converted the optimization target of the U-Net from the zeroth diffraction order in the center of diffraction field to the first diffraction order in the edge of that. And we also used a method called “phase recombination” to improve the structure of U-Net for less memory footprint and faster generating speed. Finally, the holograms with the 4K resolution have been generated in 0.05s, and the average peak signal to noise ratio (PSNR) of the reconstructed images is about 37.2 dB in DIV2K-valid-HR dataset.
Neural networks are vulnerable to meticulously crafted adversarial examples, leading to high-confidence misclassifications in image classification tasks. Due to their consistency with regular input patterns and the absence of reliance on the target model and its output information, transferable adversarial attacks exhibit a notably high stealthiness and detection difficulty, making them a significant focus of defense. In this work, we propose a deep learning defense known as multi-source adversarial perturbations elimination (MAPE) to counter diverse transferable attacks. MAPE comprises the single-source adversarial perturbation elimination (SAPE) mechanism and the pre-trained models probabilistic scheduling algorithm (PPSA). SAPE utilizes a thoughtfully designed channel-attention U-Net as the defense model and employs adversarial examples generated by a pre-trained model (e.g., ResNet) for its training, thereby enabling the elimination of known adversarial perturbations. PPSA introduces model difference quantification and negative momentum to strategically schedule multiple pre-trained models, thereby maximizing the differences among adversarial examples during the defense model’s training and enhancing its robustness in eliminating adversarial perturbations. MAPE effectively eliminates adversarial perturbations in various adversarial examples, providing a robust defense against attacks from different substitute models. In a black-box attack scenario utilizing ResNet-34 as the target model, our approach achieves average defense rates of over 95.1% on CIFAR-10 and over 71.5% on Mini-ImageNet, demonstrating state-of-the-art performance.