Multi-modal datasets in artificial intelligence (AI) often capture a third-person perspective, but our embodied human intelligence evolved with sensory input from the egocentric, first-person perspective. Towards embodied AI, we introduce the Egocentric Communications (EgoCom) dataset to advance the state-of-the-art in conversational AI, natural language, audio speech analysis, computer vision, and machine learning. EgoCom is a first-of-its-kind natural conversations dataset containing multi-modal human communication data captured simultaneously from the participants' egocentric perspectives. EgoCom includes 38.5 hours of synchronized embodied stereo audio, egocentric video with 240,000 ground-truth, time-stamped word-level transcriptions and speaker labels from 34 diverse speakers. We study baseline performance on two novel applications that benefit from embodied data: (1) predicting turn-taking in conversations and (2) multi-speaker transcription. For (1), we investigate Bayesian baselines to predict turn-taking within 5 percent of human performance. For (2), we use simultaneous egocentric capture to combine Google speech-to-text outputs, improving global transcription by 79 percent relative to a single perspective. Both applications exploit EgoCom's synchronous multi-perspective data to augment performance of embodied AI tasks.

We propose a novel approach for 3D video synthesis that is able to represent multi-view video recordings of a dynamic real-world scene in a compact, yet expressive representation that enables high-quality view synthesis and motion interpolation. Our approach takes the high quality and compactness of static neural radiance fields in a new direction: to a model-free, dynamic setting. At the core of our approach is a novel time-conditioned neural radiance field that represents scene dynamics using a set of compact latent codes. We are able to significantly boost the training speed and perceptual quality of the generated imagery by a novel hierarchical training scheme in combination with ray importance sampling. Our learned representation is highly compact and able to represent a 10 second 30 FPS multi-view video recording by 18 cameras with a model size of only 28MB. We demonstrate that our method can render high-fidelity wide-angle novel views at over 1K resolution, even for complex and dynamic scenes. We perform an extensive qualitative and quantitative evaluation that shows that our approach outperforms the state of the art. Project website: https://neural-3d-video.github.io/.

We present STaR, a novel method that performs Self-supervised Tracking and Reconstruction of dynamic scenes with rigid motion from multi-view RGB videos without any manual annotation. Recent work has shown that neural networks are surprisingly effective at the task of compressing many views of a scene into a learned function which maps from a viewing ray to an observed radiance value via volume rendering. Unfortunately, these methods lose all their predictive power once any object in the scene has moved. In this work, we explicitly model rigid motion of objects in the context of neural representations of radiance fields. We show that without any additional human specified supervision, we can reconstruct a dynamic scene with a single rigid object in motion by simultaneously decomposing it into its two constituent parts and encoding each with its own neural representation. We achieve this by jointly optimizing the parameters of two neural radiance fields and a set of rigid poses which align the two fields at each frame. On both synthetic and real world datasets, we demonstrate that our method can render photorealistic novel views, where novelty is measured on both spatial and temporal axes. Our factored representation furthermore enables animation of unseen object motion.

Object-oriented maps are important for scene understanding since they jointly capture geometry and semantics, allow individual instantiation and meaningful reasoning about objects. We introduce FroDO, a method for accurate 3D reconstruction of object instances from RGB video that infers their location, pose and shape in a coarse to fine manner. Key to FroDO is to embed object shapes in a novel learnt shape space that allows seamless switching between sparse point cloud and dense DeepSDF decoding. Given an input sequence of localized RGB frames, FroDO first aggregates 2D detections to instantiate a 3D bounding box per object. A shape code is regressed using an encoder network before optimizing shape and pose further under the learnt shape priors using sparse or dense shape representations. The optimization uses multi-view geometric, photometric and silhouette losses. We evaluate on real-world datasets, including Pix3D, Redwood-OS, and ScanNet, for single-view, multi-view, and multi-object reconstruction.

Augmented reality overlays virtual objects on the real world. To do so, the head mounted display (HMD) needs to be calibrated to establish a mapping between 3D points in the real world with 2D pixels on display panels. This distortion is a high-dimensional function that also depends on pupil position and varifocal settings. We present Raycast calibration, an efficient approach to geometrically calibrate AR displays with off-axis reflective combiners. Our approach requires a small amount of data to estimate a compact, physics-based, and ray-traceable model of the HMD optics. We apply this technique to automatically calibrate an AR prototype with display, SLAM and eye-tracker, without user in the loop.

Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction. These provide trade-offs across fidelity, efficiency and compression capabilities. In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set of the learned function while explicitly representing the classification of space as being part of the shapes interior or not. While classical SDF's both in analytical or discretized voxel form typically represent the surface of a single shape, DeepSDF can represent an entire class of shapes. Furthermore, we show state-of-the-art performance for learned 3D shape representation and completion while reducing the model size by an order of magnitude compared with previous work.

Planar reflective surfaces such as glass and mirrors are notoriously hard to reconstruct for most current 3D scanning techniques. When treated naïvely, they introduce duplicate scene structures, effectively destroying the reconstruction altogether. Our key insight is that an easy to identify structure attached to the scanner---in our case an AprilTag---can yield reliable information about the existence and the geometry of glass and mirror surfaces in a scene. We introduce a fully automatic pipeline that allows us to reconstruct the geometry and extent of planar glass and mirror surfaces while being able to distinguish between the two. Furthermore, our system can automatically segment observations of multiple reflective surfaces in a scene based on their estimated planes and locations. In the proposed setup, minimal additional hardware is needed to create high-quality results. We demonstrate this using reconstructions of several scenes with a variety of real mirrors and glass.

This paper describes a general continuous-time framework for visual-inertial simultaneous localization and mapping and calibration. We show how to use a spline parameterization that closely matches the torque-minimal motion of the sensor. Compared to traditional discrete-time solutions, the continuous-time formulation is particularly useful for solving problems with high-frame rate sensors and multiple unsynchronized devices. We demonstrate the applicability of the method for multi-sensor visual-inertial SLAM and calibration by accurately establishing the relative pose and internal parameters of multiple unsynchronized devices. We also show the advantages of the approach through evaluation and uniform treatment of both global and rolling shutter cameras within visual and visual-inertial SLAM systems.

DTAM is a system for real-time camera tracking and reconstruction which relies not on feature extraction but dense, every pixel methods. As a single hand-held RGB camera flies over a static scene, we estimate detailed textured depth maps at selected keyframes to produce a surface patchwork with millions of vertices. We use the hundreds of images available in a video stream to improve the quality of a simple photometric data term, and minimise a global spatially regularised energy functional in a novel non-convex optimisation framework. Interleaved, we track the camera's 6DOF motion precisely by frame-rate whole image alignment against the entire dense model. Our algorithms are highly parallelisable throughout and DTAM achieves real-time performance using current commodity GPU hardware. We demonstrate that a dense model permits superior tracking performance under rapid motion compared to a state of the art method using features; and also show the additional usefulness of the dense model for real-time scene interaction in a physics-enhanced augmented reality application.

As an increasing number of automatic safety and navigation features are added to modern vehicles, the crucial job of providing real-time localisation is predominantly performed by a single sensor, GPS, despite its well-known failings, particularly in urban environments. Various attempts have been made to supplement GPS to improve localisation performance, but these usually require additional specialised and expensive sensors. Offering increased value to vehicle OEMs, we show that it is possible to use just the video stream from a rear parking camera to produce smooth and locally accurate visual odometry in real-time. We use an efficient whole image alignment approach based on ESM, taking account of both the difficulties and advantages of the fact that a parking camera views only the road surface directly behind a vehicle. Visual odometry is complementary to GPS in offering localisation information at 30 Hz which is smooth and highly accurate locally whilst GPS is course but offers absolute measurements. We demonstrate our system in a large scale experiment covering real urban driving. We also present real-time fusion of our visual estimation with automotive GPS to generate a commodity-cost localisation solution which is smooth, accurate and drift free in global coordinates.