We present a novel neural algorithm for performing high-quality, highresolution, real-time novel view synthesis. From a sparse set of input RGB images or videos streams, our network both reconstructs the 3D scene and renders novel views at 1080p resolution at 30fps on an NVIDIA A100. Our feed-forward network generalizes across a wide variety of datasets and scenes and produces state-of-the-art quality for a real-time method. Our quality approaches, and in some cases surpasses, the quality of some of the top offline methods. In order to achieve these results we use a novel combination of several key concepts, and tie them together into a cohesive and effective algorithm. We build on previous works that represent the scene using semi-transparent layers and use an iterative learned render-and-refine approach to improve those layers. Instead of flat layers, our method reconstructs layered depth maps (LDMs) that efficiently represent scenes with complex depth and occlusions. The iterative update steps are embedded in a multi-scale, UNet-style architecture to perform as much compute as possible at reduced resolution. Within each update step, to better aggregate the information from multiple input views, we use a specialized Transformer-based network component. This allows the majority of the per-input image processing to be performed in the input image space, as opposed to layer space, further increasing efficiency. Finally, due to the real-time nature of our reconstruction and rendering, we dynamically create and discard the internal 3D geometry for each frame, generating the LDM for each view. Taken together, this produces a novel and effective algorithm for view synthesis. Through extensive evaluation, we demonstrate that we achieve state-of-the-art quality at real-time rates.

Casual photography is often performed in uncontrolled lighting that can result in low quality images and degrade the performance of downstream processing. We consider the problem of estimating surface normal and reflectance maps of scenes depicting people despite these conditions by supplementing the available visible illumination with a single near infrared (NIR) light source and camera, a socalled "dark flash image". Our method takes as input a single color image captured under arbitrary visible lighting and a single dark flash image captured under controlled front-lit NIR lighting at the same viewpoint, and computes a normal map, a diffuse albedo map, and a specular intensity map of the scene. Since ground truth normal and reflectance maps of faces are difficult to capture, we propose a novel training technique that combines information from two readily available and complementary sources: a stereo depth signal and photometric shading cues. We evaluate our method over a range of subjects and lighting conditions and describe two applications: optimizing stereo geometry and filling the shadows in an image.

We present a real-time bidirectional communication system that lets two people, separated by distance, experience a face-to-face conversation as if they were copresent. It is the first telepresence system that is demonstrably better than 2D videoconferencing, as measured using participant ratings (e.g., presence, attentiveness, reaction-gauging, engagement), meeting recall, and observed nonverbal behaviors (e.g., head nods, eyebrow movements). This milestone is reached by maximizing audiovisual fidelity and the sense of copresence in all design elements, including physical layout, lighting, face tracking, multi-view capture, microphone array, multi-stream compression, loudspeaker output, and lenticular display. Our system achieves key 3D audiovisual cues (stereopsis, motion parallax, and spatialized audio) and enables the full range of communication cues (eye contact, hand gestures, and body language), yet does not require special glasses or body-worn microphones/headphones. The system consists of a head-tracked autostereoscopic display, high-resolution 3D capture and rendering subsystems, and network transmission using compressed color and depth video streams. Other contributions include a novel image-based geometry fusion algorithm, free-space dereverberation, and talker localization.

We present ESPReSSo, the first real-time implementation of spacetime stereo, offering improved quality vs. existing real-time systems. ESPReSSo is a local stereo reconstruction algorithm that precomputes subpixel-shifted binary descriptors, then iteratively samples them along slanted disparity plane hypotheses, applying an edge-aware filter for spatial cost aggregation. Plane hypotheses are shared within rectangular tiles, but every pixel selects a different winner from among these candidates, as in PatchMatch Filter. This architecture performs few descriptor computations but many cost aggregations, and we tune our choice of descriptor and filter accordingly: We propose a new 32-bit binary spacetime descriptor "breve" combining the benefits of small spatial extent with robustness to scene motion, and the system aggregates costs using the permeability filter, a very efficient edge-aware filter. Our prototype outputs 60 depth frames per second on a desktop GPU, using less than 11ms total computation per frame.

Consumer time-of-flight depth cameras like Kinect and PMD are cheap, compact and produce video-rate depth maps in short-range applications. In this paper we apply energy-efficient epipolar imaging to the ToF domain to significantly expand the versatility of these sensors: we demonstrate live 3D imaging at over 15 m range outdoors in bright sunlight; robustness to global transport effects such as specular and diffuse inter-reflections---the first live demonstration for this ToF technology; interference-free 3D imaging in the presence of many ToF sensors, even when they are all operating at the same optical wavelength and modulation frequency; and blur-free, distortion-free 3D video in the presence of severe camera shake. We believe these achievements can make such cheap ToF devices broadly applicable in consumer and robotics domains.

Programmable coding of light between a source and a sensor has led to several important results in computational illumination, imaging and display. Little is known, however, about how to utilize energy most effectively, especially for applications in live imaging. In this paper, we derive a novel framework to maximize energy efficiency by "homogeneous matrix factorization" that respects the physical constraints of many coding mechanisms (DMDs/LCDs, lasers, etc. ). We demonstrate energy-efficient imaging using two prototypes based on DMD and laser illumination. For our DMD-based prototype, we use fast local optimization to derive codes that yield brighter images with fewer artifacts in many transport probing tasks. Our second prototype uses a novel combination of a low-power laser projector and a rolling shutter camera. We use this prototype to demonstrate never-seen-before capabilities such as (1) capturing live structured-light video of very bright scenes---even a light bulb that has been turned on; (2) capturing epipolar-only and indirect-only live video with optimal energy efficiency; (3) using a low-power projector to reconstruct 3D objects in challenging conditions such as strong indirect light, strong ambient light, and smoke; and (4) recording live video from a projector's---rather than the camera's---point of view.

Separating the direct and global components of radiance can aid shape recovery algorithms and can provide useful information about materials in a scene. Practical methods for finding the direct and global components use multiple images captured under varying illumination patterns and require the scene, light source and camera to remain stationary during the image acquisition process. In this paper, we develop a motion compensation method that relaxes this condition and allows direct-global separation to be performed on video sequences of dynamic scenes captured by moving projector-camera systems. Key to our method is being able to register frames in a video sequence to each other in the presence of time varying, high frequency active illumination patterns. We compare our motion compensated method to alternatives such as single shot separation and frame interleaving as well as ground truth. We present results on challenging video sequences that include various types of motions and deformations in scenes that contain complex materials like fabric, skin, leaves and wax.

This paper introduces a vision based localization method for large scale urban environments. The method is based upon Bag-of-Words image retrieval techniques and handles problems that arise in urban environments due to repetitive scene structure and the presence of dynamic objects like vehicles. The localization system was experimentally verified it localization experiments along a 5km long path in an urban environment.

Here we consider the problem of automatically segmenting images taken from a boat or low-flying aircraft. Such a capability is important for autonomous river following and mapping. The need for accurate segmentation in a wide variety of riverine environments challenges the state of the art vision-based methods that have been used in more structured environments such as roads and highways. Apart from the lack of structure, the principal difficulty is the large spatial and temporal variations in the appearance of water in the presence of nearby vegetation and with reflections from the sky. We propose a self-supervised method to segment images into `sky', `river' and `shore' (vegetation + structures) regions. Our approach uses assumptions about river scene structure to learn appearance models based on features like color, texture and image location which are used to segment the image. We validated our algorithm by testing on four datasets captured under varying conditions on different rivers. Our self-supervised algorithm had higher accuracy rates than a supervised alternative, often significantly more accurate, and does not need to be retrained to work under different conditions.

We consider the perceptual challenges inherent in the robotic manipulation of previously unseen socks, with the end goal of manipulation by a household robot for laundry. The task poses challenging problems in modeling the appearance, shape and configuration of these textile items that tend to exhibit high variability in texture, design, and style while being highly articulated objects.

The harvest yield in vineyards can vary significantly from year to year and also spatially within plots due to variations in climate, soil conditions and pests. Fine grained knowledge of crop yields can allow viticulturists to better manage their vineyards. The current industry practice for yield prediction is destructive, expensive and spatially sparse - during the growing season sparse samples are taken and extrapolated to determine overall yield. We present an automated method that uses computer vision to detect and count grape berries. The method could potentially be deployed across large vineyards taking measurements at every vine in a non-destructive manner. Our berry detection uses both shape and visual texture and we can demonstrate detection of green berries against a green leaf background. Berry detections are counted and the eventual harvest yield is predicted. Results are presented for 224 vines (over 450 meters) of two different grape varieties and compared against the actual harvest yield as groundtruth. We calibrate our berry count to yield and find that we can predict yield of individual vineyard rows to within 9.8% of actual crop weight.

In this paper we present a novel approach to robust visual servoing. This method removes the feature tracking step from a typical visual servoing algorithm. We do not need correspondences of the features for deriving the control signal. This is achieved by modeling the image features as a Mixture of Gaussians in the current as well as desired images. Using Lyapunov theory, a control signal is derived to minimize a distance function between the two Gaussian mixtures. The distance function is given in a closed form, and its gradient can be efficiently computed and used to control the system. For simplicity, we first consider the 2D motion case. Then, the general case is presented by introducing the depth distribution of the features to control the six degrees of freedom. Experiments are conducted within a simulation framework to validate our proposed method.

Image-based navigation paradigms have recently emerged as an interesting alternative to conventional model- based methods in mobile robotics. In this paper, we augment the existing image-based navigation approaches by presenting a novel image-based exploration algorithm. The algorithm facilitates a mobile robot equipped only with a monocular pan-tilt camera to autonomously explore a typical indoor environment. The algorithm infers frontier information directly from the images and displaces the robot towards regions that are informative for navigation. The frontiers are detected using a geometric context-based segmentation scheme that exploits the natural scene structure in indoor environments. In the due process, a topological graph of the workspace is built in terms of images which can be subsequently utilised for the tasks of localisation, path planning and navigation. Experimental results on a mobile robot in an unmodified laboratory and corridor environments demonstrate the validity of the approach.