Learning editable high-resolution scene representations for dynamic scenes is an open problem with applications across the domains from autonomous driving to creative editing - the most successful approaches today make a trade-off between editability and supporting scene complexity: neural atlases represent dynamic scenes as two deforming image layers, foreground and background, which are editable in 2D, but break down when multiple objects occlude and interact. In contrast, scene graph models make use of annotated data such as masks and bounding boxes from autonomous-driving datasets to capture complex 3D spatial relationships, but their implicit volumetric node representations are challenging to edit view-consistently. We propose Neural Atlas Graphs (NAGs), a hybrid high-resolution scene representation, where every graph node is a view-dependent neural atlas, facilitating both 2D appearance editing and 3D ordering and positioning of scene elements. Fit at test-time, NAGs achieve state-of-the-art quantitative results on the Waymo Open Dataset - by 5 dB PSNR increase compared to existing methods - and make environmental editing possible in high resolution and visual quality - creating counterfactual driving scenarios with new backgrounds and edited vehicle appearance. We find that the method also generalizes beyond driving scenes and compares favorably - by more than 7 dB in PSNR - to recent matting and video editing baselines on the DAVIS video dataset with a diverse set of human and animal-centric scenes. Project Page: https://princeton-computational-imaging.github.io/nag/

Each photo in an image burst can be considered a sam-ple of a complex 3D scene: the product of parallax, diffuse and specular materials, scene motion, and illuminant vari-ation. While decomposing all of these effects from a stack of misaligned images is a highly ill-conditioned task, the conventional align-and-merge burst pipeline takes the other extreme: blending them into a single image. In this work, we propose a versatile intermediate representation: a two-layer alpha-composited image plus flow model constructed with neural spline fields - networks trained to map input coordinates to spline control points. Our method is able to, during test-time optimization, jointly fuse a burst image capture into one high-resolution reconstruction and decom-pose it into transmission and obstruction layers. Then, by discarding the obstruction layer, we can perform a range of tasks including seeing through occlusions, reflection sup-pression, and shadow removal. Tested on complex in-the-wild captures we find that, with no post-processing steps or learned priors, our generalizable model is able to out-perform existing dedicated single-image and multi-view ob-struction removal approaches.

Challenging to capture, and challenging to display on a cellphone screen, the panorama paradoxically remains both a staple and underused feature of modern mobile camera applications. In this work we address both of these challenges with a spherical neural light field model for implicit panoramic image stitching and re-rendering; able to accommodate for depth parallax, view-dependent lighting, and local scene motion and color changes during capture. Fit during test-time to an arbitrary path panoramic video capture -- vertical, horizontal, random-walk -- these neural light spheres jointly estimate the camera path and a high-resolution scene reconstruction to produce novel wide field-of-view projections of the environment. Our single-layer model avoids expensive volumetric sampling, and decomposes the scene into compact view-dependent ray offset and color components, with a total model size of 80 MB per scene, and real-time (50 FPS) rendering at 1080p resolution. We demonstrate improved reconstruction quality over traditional image stitching and radiance field methods, with significantly higher tolerance to scene motion and non-ideal capture settings.

While conventional cameras offer versatility for applications ranging from amateur photography to autonomous driving, computational cameras allow for domain-specific adaption. Cameras with co-designed optics and image processing algorithms enable high-dynamic-range image recovery, depth estimation, and hyperspectral imaging through optically encoding scene information that is otherwise undetected by conventional cameras. However, this optical encoding creates a challenging inverse reconstruction problem for conventional image recovery, and often lowers the overall photographic quality. Thus computational cameras with domain-specific optics have only been adopted in a few specialized applications where the captured information cannot be acquired in other ways. In this work, we investigate a method that combines two optical systems into one to tackle this challenge. We split the aperture of a conventional camera into two halves: one which applies an application-specific modulation to the incident light via a diffractive optical element to produce a coded image capture, and one which applies no modulation to produce a conventional image capture. Co-designing the phase modulation of the split aperture with a dual-pixel sensor allows us to simultaneously capture these coded and uncoded images without increasing physical or computational footprint. With an uncoded conventional image alongside the optically coded image in hand, we investigate image reconstruction methods that are conditioned on the conventional image, making it possible to eliminate artifacts and compute costs that existing methods struggle with. We assess the proposed method with 2-in-1 cameras for optical high-dynamic-range reconstruction, monocular depth estimation, and hyperspectral imaging, comparing favorably to all tested methods in all applications.

Modern mobile burst photography pipelines capture and merge a short sequence of frames to recover an enhanced image, but often disregard the 3D nature of the scene they capture, treating pixel motion between images as a 2D aggregation problem. We show that in a “long-burst”, forty-two 12-megapixel RAW frames captured in a two-second sequence, there is enough parallax information from natural hand tremor alone to recover high-quality scene depth. To this end, we devise a test-time optimization approach that fits a neural RGB-D representation to long-burst data and simultaneously estimates scene depth and camera motion. Our plane plus depth model is trained end-to-end, and performs coarse-to-fine refinement by controlling which multi-resolution volume features the network has access to at what time during training. We validate the method experimentally, and demonstrate geometrically accurate depth reconstructions with no additional hardware or separate data pre-processing and pose-estimation steps.

Depth sensors have emerged as a cornerstone sensor modality with diverse applications in personal hand-held devices, robotics, scientific imaging, autonomous vehicles, and more. In particular, correlation Time-of-Flight (ToF) sensors have found widespread adoption for meter-scale indoor applications such as object tracking and pose estimation. While they offer high depth resolution at competitive costs, the precision of these indirect ToF sensors is fundamentally limited by their modulation contrast, which is in turn limited by the effects of photo-conversion noise. In contrast, optical interferometric methods can leverage short illumination modulation wavelengths to achieve depth precision three orders of magnitude greater than ToF, but typically find their range is restricted to the sub-centimeter. In this work, we merge concepts from both correlation ToF design and interferometric imaging; a step towards bridging the gap between these methods. We propose a computational ToF imaging method that optically computes the GHz ToF correlation signal in free space before photo-conversion. To acquire a depth map, we scan a scene point-wise and computationally unwrap the collected correlation measurements. Specifically, we repurpose electro-optical modulators used in optical communication for ToF imaging with centimeter-wave signals, and achieve all-optical correlation at 7.15 GHz and 14.32 GHz modulation frequencies. While GHz modulation frequencies increase depth precision, these high modulation rates also pose a technical challenge. They result in dozens of wraps per meter which cannot be estimated robustly by existing phase unwrapping methods. We tackle this problem with a proposed segmentation-inspired phase unwrapping network , which exploits the correlation of adjacent GHz phase measurements to classify regions into their respective wrap counts. We validate this method in simulation and experimentally, and demonstrate precise depth sensing using centimeter wave modulation that is robust to surface texture and ambient light. Compared to existing analog demodulation methods, the proposed system outperforms all of them across all tested scenarios. CCS Concepts: • Computing methodologies ;

Modern smartphones can continuously stream multi-megapixel RGB images at 60 Hz, synchronized with high-quality 3D pose information and low-resolution LiDAR-driven depth estimates. During a snapshot photograph, the natural unsteadiness of the photographer's hands offers millimeter-scale variation in camera pose, which we can capture along with RGB and depth in a circular buffer. In this work we explore how, from a bundle of these measurements acquired during viewfinding, we can combine dense micro-baseline parallax cues with kilopixel LiDAR depth to distill a high-fidelity depth map. We take a test-time optimization approach and train a coordinate MLP to output photometrically and geometrically consistent depth estimates at the continuous coordinates along the path traced by the photographer's natural hand shake. With no additional hardware, artificial hand motion, or user interaction beyond the press of a button, our proposed method brings high-resolution depth estimates to point-and-shoot “table-top” photography – textured objects at close range.

We investigate the generalization capabilities of neural signed distance functions (SDFs) for learning 3D object representations for unseen and unlabeled point clouds. Existing methods can fit SDFs to a handful of object classes and boast fine detail or fast inference speeds, but do not generalize well to unseen shapes. We introduce a two-stage semi-supervised meta-learning approach that transfers shape priors from labeled to unlabeled data to reconstruct unseen object categories. The first stage uses an episodic training scheme to simulate training on unlabeled data and meta-learns initial shape priors. The second stage then introduces unlabeled data with disjoint classes in a semi-supervised scheme to diversify these priors and achieve generalization. We assess our method on both synthetic data and real collected point clouds. Experimental results and analysis validate that our approach outperforms existing neural SDF methods and is capable of robust zero-shot inference on 100+ unseen classes. Code can be found at https://github.com/princeton-computational-imaging/gensdf

Static gesture recognition is an effective non-verbal communication channel between a user and their devices; however many modern methods are sensitive to the relative pose of the user's hands with respect to the capture device, as parts of the gesture can become occluded. We present two methodologies for gesture recognition via synchronized recording from two depth cameras to alleviate this occlusion problem. One is a more classic approach using iterative closest point registration to accurately fuse point clouds and a single PointNet architecture for classification, and the other is a dual Point-Net architecture for classification without registration. On a manually collected data-set of 20,100 point clouds we show a 39.2% reduction in misclassification for the fused point cloud method, and 53.4% for the dual PointNet, when compared to a standard single camera pipeline.