Neural scene representations, both continuous and discrete, have recently emerged as a powerful new paradigm for 3D scene understanding. Recent efforts have tackled unsupervised discovery of object-centric neural scene representations. However, the high cost of ray-marching, exacerbated by the fact that each object representation has to be ray-marched separately, leads to insufficiently sampled radiance fields and thus, noisy renderings, poor framerates, and high memory and time complexity during training and rendering. Here, we propose to represent objects in an object-centric, compositional scene representation as light fields. We propose a novel light field compositor module that enables reconstructing the global light field from a set of object-centric light fields. Dubbed Compositional Object Light Fields (COLF), our method enables unsupervised learning of object-centric neural scene representations, state-of-the-art reconstruction and novel view synthesis performance on standard datasets, and rendering and training speeds at orders of magnitude faster than existing 3D approaches.

Reconstruction of 3D neural fields from posed images has emerged as a promising method for self-supervised representation learning. The key challenge preventing the deployment of these 3D scene learners on large-scale video data is their dependence on precise camera poses from structure-from-motion, which is prohibitively expensive to run at scale. We propose a method that jointly reconstructs camera poses and 3D neural scene representations online and in a single forward pass. We estimate poses by first lifting frame-to-frame optical flow to 3D scene flow via differentiable rendering, preserving locality and shift-equivariance of the image processing backbone. SE(3) camera pose estimation is then performed via a weighted least-squares fit to the scene flow field. This formulation enables us to jointly supervise pose estimation and a generalizable neural scene representation via re-rendering the input video, and thus, train end-to-end and fully self-supervised on real-world video datasets. We demonstrate that our method performs robustly on diverse, real-world video, notably on sequences traditionally challenging to optimization-based pose estimation techniques.

We introduce a method for novel view synthesis given only a single wide-baseline stereo image pair. In this challenging regime, 3D scene points are regularly observed only once, requiring prior-based reconstruction of scene geometry and appearance. We find that existing approaches to novel view synthesis from sparse observations fail due to recovering incorrect 3D geometry and due to the high cost of differentiable rendering that precludes their scaling to large-scale training. We take a step towards resolving these shortcomings by formulating a multi-view transformer encoder, proposing an efficient, image-space epipolar line sampling scheme to assemble image features for a target ray, and a lightweight cross-attention-based renderer. Our contributions enable training of our method on a large-scale real-world dataset of indoor and outdoor scenes. We demonstrate that our method learns powerful multi-view geometry priors while reducing the rendering time. We conduct extensive comparisons on held-out test scenes across two real-world datasets, significantly outperforming prior work on novel view synthesis from sparse image observations and achieving multi-view-consistent novel view synthesis.

We present a method to map 2D image observations of a scene to a persistent 3D scene representation, enabling novel view synthesis and disentangled representation of the movable and immovable components of the scene. Motivated by the bird's-eye-view (BEV) representation commonly used in vision and robotics, we propose conditional neural groundplans, ground-aligned 2D feature grids, as persistent and memory-efficient scene representations. Our method is trained selfsupervised from unlabeled multi-view observations using differentiable rendering, and learns to complete geometry and appearance of occluded regions. In addition, we show that we can leverage multi-view videos at training time to learn to separately reconstruct static and movable components of the scene from a single image at test time. The ability to separately reconstruct movable objects enables a variety of downstream tasks using simple heuristics, such as extraction of objectcentric 3D representations, novel view synthesis, instance-level segmentation, 3D bounding box prediction, and scene editing. This highlights the value of neural groundplans as a backbone for efficient 3D scene understanding models. We study the problem of inferring a persistent 3D scene representation given a few image observations, while disentangling static scene components from movable objects (referred to as dynamic). Recent works in differentiable rendering have made significant progress in the long-standing problem of 3D reconstruction from small sets of image observations (Yu et al., 2020; Sitzmann et al., 2019b; Sajjadi et al., 2021). Approaches based on pixel-aligned features (Yu et al., 2020; Trevithick & Yang, 2021; Henzler et al., 2021) have achieved plausible novel view synthesis of scenes composed of independent objects from single images. However, these methods do not produce persistent 3D scene representations that can be directly processed in 3D, for instance, via 3D convolutions. Instead, all processing has to be performed in image space. In contrast, some methods infer 3D voxel grids, enabling processing such as geometry and appearance completion via shift-equivariant 3D convolutions (Lal et al., 2021; Guo et al., 2022), which is however expensive both in terms of computation and memory. Meanwhile, bird's-eye-view (BEV) representations, 2D grids aligned with the ground plane of a scene, have been fruitfully deployed as state representations for navigation, layout generation, and future frame prediction (Saha et al., 2022; Philion & Fidler, 2020; Roddick et al., 2019; Jeong et al., 2022; Mani et al., 2020). While they compress the height axis and are thus not a full 3D representation, 2D convolutions on top of BEVs retain shift-equivariance in the ground plane and are, in contrast to image-space convolutions, free of perspective camera distortions.

We present Neural Descriptor Fields (NDFs), an object representation that encodes both points and relative poses between an object and a target (such as a robot gripper or a rack used for hanging) via category-level descriptors. We employ this representation for object manipulation, where given a task demonstration, we want to repeat the same task on a new object instance from the same category. We propose to achieve this objective by searching (via optimization) for the pose whose descriptor matches that observed in the demonstration. NDFs are conveniently trained in a self-supervised fashion via a 3D auto-encoding task that does not rely on expert-labeled keypoints. Further, NDFs are SE(3)-equivariant, guaranteeing performance that generalizes across all possible 3D object translations and rotations. We demonstrate learning of manipulation tasks from few (5-10) demonstrations both in simulation and on a real robot. Our performance generalizes across both object instances and 6-DoF object poses, and significantly outperforms a recent baseline that relies on 2D descriptors. Project website: https://yilundu.github.io/ndf/.

Deep neural networks have been used widely to learn the latent structure of datasets, across modalities such as images, shapes, and audio signals. However, existing models are generally modality-dependent, requiring custom architectures and objectives to process different classes of signals. We leverage neural fields to capture the underlying structure in image, shape, audio and cross-modal audiovisual domains in a modality-independent manner. We cast our task as one of learning a manifold, where we aim to infer a low-dimensional, locally linear subspace in which our data resides. By enforcing coverage of the manifold, local linearity, and local isometry, our model -- dubbed GEM -- learns to capture the underlying structure of datasets across modalities. We can then travel along linear regions of our manifold to obtain perceptually consistent interpolations between samples, and can further use GEM to recover points on our manifold and glean not only diverse completions of input images, but cross-modal hallucinations of audio or image signals. Finally, we show that by walking across the underlying manifold of GEM, we may generate new samples in our signal domains. Code and additional results are available at https://yilundu.github.io/gem/.

Inferring representations of 3D scenes from 2D observations is a fundamental problem of computer graphics, computer vision, and artificial intelligence. Emerging 3D-structured neural scene representations are a promising approach to 3D scene understanding. In this work, we propose a novel neural scene representation, Light Field Networks or LFNs, which represent both geometry and appearance of the underlying 3D scene in a 360-degree, four-dimensional light field parameterized via a neural implicit representation. Rendering a ray from an LFN requires only a *single* network evaluation, as opposed to hundreds of evaluations per ray for ray-marching or volumetric based renderers in 3D-structured neural scene representations. In the setting of simple scenes, we leverage meta-learning to learn a prior over LFNs that enables multi-view consistent light field reconstruction from as little as a single image observation. This results in dramatic reductions in time and memory complexity, and enables real-time rendering. The cost of storing a 360-degree light field via an LFN is two orders of magnitude lower than conventional methods such as the Lumigraph. Utilizing the analytical differentiability of neural implicit representations and a novel parameterization of light space, we further demonstrate the extraction of sparse depth maps from LFNs.

The advent of deep learning has given rise to neural scene representations - learned mathematical models of a 3D environment. However, many of these representations do not explicitly reason about geometry and thus do not account for the underlying 3D structure of the scene. In contrast, geometric deep learning has explored 3D-structure-aware representations of scene geometry, but requires explicit 3D supervision. We propose Scene Representation Networks (SRNs), a continuous, 3D-structure-aware scene representation that encodes both geometry and appearance. SRNs represent scenes as continuous functions that map world coordinates to a feature representation of local scene properties. By formulating the image formation as a differentiable ray-marching algorithm, SRNs can be trained end-to-end from only 2D observations, without access to depth or geometry. This formulation naturally generalizes across scenes, learning powerful geometry and appearance priors in the process. We demonstrate the potential of SRNs by evaluating them for novel view synthesis, few-shot reconstruction, joint shape and appearance interpolation, and unsupervised discovery of a non-rigid face model.