Abstract:The ability to present three-dimensional (3D) scenes with continuous depth sensation has a profound impact on virtual and augmented reality, human–computer interaction, education and training. Computer-generated holography (CGH) enables high-spatio-angular-resolution 3D projection via numerical simulation of diffraction and interference<a href="/articles/s41586-020-03152-0#ref-CR1">1</a>. Yet, existing physically based methods fail to produce holograms with both per-pixel focal control and accurate occlusion<a href="/articles/s41586-020-03152-0#ref-CR2">2</a>,<a href="/articles/s41586-020-03152-0#ref-CR3">3</a>. The computationally taxing Fresnel diffraction simulation further places an explicit trade-off between image quality and runtime, making dynamic holography impractical<a href="/articles/s41586-020-03152-0#ref-CR4">4</a>. Here we demonstrate a deep-learning-based CGH pipeline capable of synthesizing a photorealistic colour 3D hologram from a single RGB-depth image in real time. Our convolutional neural network (CNN) is extremely memory efficient (below 620 kilobytes) and runs at 60 hertz for a resolution of 1,920 × 1,080 pixels on a single consumer-grade graphics processing unit. Leveraging low-power on-device artificial intelligence acceleration chips, our CNN also runs interactively on mobile (iPhone 11 Pro at 1.1 hertz) and edge (Google Edge TPU at 2.0 hertz) devices, promising real-time performance in future-generation virtual and augmented-reality mobile headsets. We enable this pipeline by introducing a large-scale CGH dataset (MIT-CGH-4K) with 4,000 pairs of RGB-depth images and corresponding 3D holograms. Our CNN is trained with differentiable wave-based loss functions<a href="/articles/s41586-020-03152-0#ref-CR5">5</a> and physically approximates Fresnel diffraction. With an anti-aliasing phase-only encoding method, we experimentally demonstrate speckle-free, natural-looking, high-resolution 3D holograms. Our learning-based approach and the Fresnel hologram dataset will help to unlock the full potential of holography and enable applications in metasurface design<a href="/articles/s41586-020-03152-0#ref-CR6">6</a>,<a href="/articles/s41586-020-03152-0#ref-CR7">7</a>, optical and acoustic tweezer-based microscopic manipulation<a href="#ref-CR8">8</a>,<a href="#ref-CR9">9</a>,<a href="/articles/s41586-020-03152-0#ref-CR10">10</a>, holographic microscopy<a href="/articles/s41586-020-03152-0#ref-CR11">11</a> and single-exposure volumetric 3D printing<a href="/articles/s41586-020-03152-0#ref-CR12">12</a>,<a href="/articles/s41586-020-03152-0#ref-CR13">13</a>.

Bright-field holography: cross-modality deep learning enables snapshot 3D imaging with bright-field contrast using a single hologram

Cross-modality deep learning brings bright-field microscopy contrast to holography

Deep learning-based color holographic microscopy

Unsupervised Deep Learning Enables 3D Imaging for Single‐Shot Incoherent Holography

Deep-3D Microscope: 3D volumetric microscopy of thick scattering samples using a wide-field microscope and machine learning

OAH-Net: A Deep Neural Network for Hologram Reconstruction of Off-axis Digital Holographic Microscope

Generating Real-Scene Hologram Through Light Field Imaging and Deep Learning

Extended Depth-of-field in Holographic Imaging Using Deep-Learning-based Autofocusing and Phase Recovery

Towards real-time photorealistic 3D holography with deep neural networks

Sparse deep computer-generated holography for optical microscopy

Auto-focusing and extended depth-of-field holographic reconstruction using deep learning (Conference Presentation)

Generalizable Deep Learning Approach for 3D Particle Imaging using Holographic Microscopy

Three-Dimensional Virtual Optical Clearing with Cycle-Consistent Generative Adversarial Network

High-throughput, high-resolution registration-free generated adversarial network microscopy

Generative adversarial neural network for 3D-hologram reconstruction

Single-shot deep-learning based 3D imaging of Fresnel incoherent correlation holography

Reusability report: Unpaired deep-learning approaches for holographic image reconstruction

Speeding up reconstruction of 3D tomograms in holographic flow cytometry via deep learning

GANscan: continuous scanning microscopy using deep learning deblurring

Generating Multi‐Depth 3D Holograms Using a Fully Convolutional Neural Network

Deep-learning-aided multi-focal hologram generation