The explosive growth of computation and energy cost of artificial intelligence has spurred strong interests in new computing modalities as potential alternatives to conventional electronic processors. Photonic processors that execute operations using photons instead of electrons, have promised to enable optical neural networks with ultra-low latency and power consumption. However, existing optical neural networks, limited by the underlying network designs, have achieved image recognition accuracy much lower than state-of-the-art electronic neural networks. In this work, we close this gap by introducing a large-kernel spatially-varying convolutional neural network learned via low-dimensional reparameterization techniques. We experimentally instantiate the network with a flat meta-optical system that encompasses an array of nanophotonic structures designed to induce angle-dependent responses. Combined with an extremely lightweight electronic backend with approximately 2K parameters we demonstrate a nanophotonic neural network reaches 73.80\% blind test classification accuracy on CIFAR-10 dataset, and, as such, the first time, an optical neural network outperforms the first modern digital neural network -- AlexNet (72.64\%) with 57M parameters, bringing optical neural network into modern deep learning era.
Today's commodity camera systems rely on compound optics to map light originating from the scene to positions on the sensor where it gets recorded as an image. To record images without optical aberrations, i.e., deviations from Gauss' linear model of optics, typical lens systems introduce increasingly complex stacks of optical elements which are responsible for the height of existing commodity cameras. In this work, we investigate \emph{flat nanophotonic computational cameras} as an alternative that employs an array of skewed lenslets and a learned reconstruction approach. The optical array is embedded on a metasurface that, at 700~nm height, is flat and sits on the sensor cover glass at 2.5~mm focal distance from the sensor. To tackle the highly chromatic response of a metasurface and design the array over the entire sensor, we propose a differentiable optimization method that continuously samples over the visible spectrum and factorizes the optical modulation for different incident fields into individual lenses. We reconstruct a megapixel image from our flat imager with a \emph{learned probabilistic reconstruction} method that employs a generative diffusion model to sample an implicit prior. To tackle \emph{scene-dependent aberrations in broadband}, we propose a method for acquiring paired captured training data in varying illumination conditions. We assess the proposed flat camera design in simulation and with an experimental prototype, validating that the method is capable of recovering images from diverse scenes in broadband with a single nanophotonic layer.
Light's ability to perform massive linear operations parallelly has recently inspired numerous demonstrations of optics-assisted artificial neural networks (ANN). However, a clear advantage of optics over purely digital ANN in a system-level has not yet been established. While linear operations can indeed be optically performed very efficiently, the lack of nonlinearity and signal regeneration require high-power, low-latency signal transduction between optics and electronics. Additionally, a large power is needed for the lasers and photodetectors, which are often neglected in the calculation of energy consumption. Here, instead of mapping traditional digital operations to optics, we co-optimized a hybrid optical-digital ANN, that operates on incoherent light, and thus amenable to operations under ambient light. Keeping the latency and power constant between purely digital ANN and hybrid optical-digital ANN, we identified a low-power/ latency regime, where an optical encoder provides higher classification accuracy than a purely digital ANN. However, in that regime, the overall classification accuracy is lower than what is achievable with higher power and latency. Our results indicate that optics can be advantageous over digital ANN in applications, where the overall performance of the ANN can be relaxed to prioritize lower power and latency.
The scanning fiber endoscope (SFE), an ultra-small optical imaging device with a large field-of-view (FOV) for having a clear forward view into the interior of blood vessels, has great potential in the cardio-vascular disease diagnosis and surgery assistance, which is one of the key applications for short-wave infrared (SWIR) biomedical imaging. The state-of-the-art SFE system uses a miniaturized refractive spherical lens doublet for beam projection. A meta-lens is a promising alternative which can be made much thinner and has fewer off-axis aberrations than its refractive counterpart. We report an SFE system with meta-lens working at 1310nm to achieve a resolution ($\sim 140\mu m$ at the center of field and the imaging distance of $15mm$), FOV ($\sim 70 \circ$), and depth-of-focus (DOF $\sim 15mm$), which are comparable to a state-of-the-art refractive lens SFE. The use of the meta-lens reduces the length of the optical track from $1.2mm$ to $0.86mm$. The resolution of our meta-lens based SFE drops by less than a factor of $2$ at the edge of the FOV, while the refractive lens counterpart has a $\sim 3$ times resolution degradation. These results show the promise of integrating a meta-lens into an endoscope for device minimization and optical performance improvement.
Foveated imaging provides a better tradeoff between situational awareness (field of view) and resolution and is critical in long-wavelength infrared regimes because of the size, weight, power, and cost of thermal sensors. We demonstrate computational foveated imaging by exploiting the ability of a meta-optical frontend to discriminate between different polarization states and a computational backend to reconstruct the captured image/video. The frontend is a three-element optic: the first element which we call the "foveal" element is a metalens that focuses s-polarized light at a distance of $f_1$ without affecting the p-polarized light; the second element which we call the "perifoveal" element is another metalens that focuses p-polarized light at a distance of $f_2$ without affecting the s-polarized light. The third element is a freely rotating polarizer that dynamically changes the mixing ratios between the two polarization states. Both the foveal element (focal length = 150mm; diameter = 75mm), and the perifoveal element (focal length = 25mm; diameter = 25mm) were fabricated as polarization-sensitive, all-silicon, meta surfaces resulting in a large-aperture, 1:6 foveal expansion, thermal imaging capability. A computational backend then utilizes a deep image prior to separate the resultant multiplexed image or video into a foveated image consisting of a high-resolution center and a lower-resolution large field of view context. We build a first-of-its-kind prototype system and demonstrate 12 frames per second real-time, thermal, foveated image, and video capture in the wild.
Meta-optics have rapidly become a major research field within the optics and photonics community, strongly driven by the seemingly limitless opportunities made possible by controlling optical wavefronts through interaction with arrays of sub-wavelength scatterers. As more and more modalities are explored, the design strategies to achieve desired functionalities become increasingly demanding, necessitating more advanced design techniques. Herein, the inverse-design approach is utilized to create a set of single-layer meta-optics that simultaneously focus light and shape the spectra of focused light without using any filters. Thus, both spatial and spectral properties of the meta-optics are optimized, resulting in spectra that mimic the color matching functions of the CIE 1931 XYZ color space, which links the distributions of wavelengths in light and the color perception of a human eye. Experimental demonstrations of these meta-optics show qualitative agreement with the theoretical predictions and help elucidate the focusing mechanism of these devices.
In recent years, Convolutional Neural Networks (CNNs) have enabled ubiquitous image processing applications. As such, CNNs require fast runtime (forward propagation) to process high-resolution visual streams in real time. This is still a challenging task even with state-of-the-art graphics and tensor processing units. The bottleneck in computational efficiency primarily occurs in the convolutional layers. Performing operations in the Fourier domain is a promising way to accelerate forward propagation since it transforms convolutions into elementwise multiplications, which are considerably faster to compute for large kernels. Furthermore, such computation could be implemented using an optical 4f system with orders of magnitude faster operation. However, a major challenge in using this spectral approach, as well as in an optical implementation of CNNs, is the inclusion of a nonlinearity between each convolutional layer, without which CNN performance drops dramatically. Here, we propose a Spectral CNN Linear Counterpart (SCLC) network architecture and develop a Knowledge Distillation (KD) approach to circumvent the need for a nonlinearity and successfully train such networks. While the KD approach is known in machine learning as an effective process for network pruning, we adapt the approach to transfer the knowledge from a nonlinear network (teacher) to a linear counterpart (student). We show that the KD approach can achieve performance that easily surpasses the standard linear version of a CNN and could approach the performance of the nonlinear network. Our simulations show that the possibility of increasing the resolution of the input image allows our proposed 4f optical linear network to perform more efficiently than a nonlinear network with the same accuracy on two fundamental image processing tasks: (i) object classification and (ii) semantic segmentation.
The parallelism of optics and the miniaturization of optical components using nanophotonic structures, such as metasurfaces present a compelling alternative to electronic implementations of convolutional neural networks. The lack of a low-power optical nonlinearity, however, requires slow and energy-inefficient conversions between the electronic and optical domains. Here, we design an architecture which utilizes a single electrical to optical conversion by designing a free-space optical frontend unit that implements the linear operations of the first layer with the subsequent layers realized electronically. Speed and power analysis of the architecture indicates that the hybrid photonic-electronic architecture outperforms sole electronic architecture for large image sizes and kernels. Benchmarking of the photonic-electronic architecture on a modified version of AlexNet achieves a classification accuracy of 87% on images from the Kaggle Cats and Dogs challenge database.