NeurIPS 2024 Papers

LiDAR-based 3D object detection has made impressive progress recently, yet most existing models are black-box, lacking interpretability. Previous explanation approaches primarily focus on analyzing image-based models and are not readily applicable to LiDAR-based 3D detectors. In this paper, we propose a feature factorization activation map (FFAM) to generate high-quality visual explanations for 3D detectors. FFAM employs non-negative matrix factorization to generate concept activation maps and subsequently aggregates these maps to obtain a global visual explanation. To achieve object-specific visual explanations, we refine the global visual explanation using the feature gradient of a target object. Additionally, we introduce a voxel upsampling strategy to align the scale between the activation map and input point cloud. We qualitatively and quantitatively analyze FFAM with multiple detectors on several datasets. Experimental results validate the high-quality visual explanations produced by FFAM. The code is available at \url{https://anonymous.4open.science/r/FFAM-B9AF}.

Reconstructing accurate 3D surfaces for street-view scenarios is crucial for applications such as digital entertainment and autonomous driving simulation. However, existing street-view datasets, including KITTI, Waymo, and nuScenes, only offer noisy LiDAR points as ground-truth data for geometric evaluation of reconstructed surfaces. These geometric ground-truths often lack the necessary precision to evaluate surface positions and do not provide data for assessing surface normals. To overcome these challenges, we introduce the SS3DM dataset, comprising precise \textbf{S}ynthetic \textbf{S}treet-view \textbf{3D} \textbf{M}esh models exported from the CARLA simulator. These mesh models facilitate accurate position evaluation and include normal vectors for evaluating surface normal. To simulate the input data in realistic driving scenarios for 3D reconstruction, we virtually drive a vehicle equipped with six RGB cameras and five LiDAR sensors in diverse outdoor scenes. Leveraging this dataset, we establish a benchmark for state-of-the-art surface reconstruction methods, providing a comprehensive evaluation of the associated challenges. For more information, visit our homepage at https://ss3dm.top.

It is vital to infer a signed distance function (SDF) for multi-view based surface reconstruction. 3D Gaussian splatting (3DGS) provides a novel perspective for volume rendering, and shows advantages in rendering efficiency and quality. Although 3DGS provides a promising neural rendering option, it is still hard to infer SDFs for surface reconstruction with 3DGS due to the discreteness, the sparseness, and the off-surface drift of 3D Gaussians. To resolve these issues, we propose a method that seamlessly merge 3DGS with the learning of neural SDFs. Our key idea is to more effectively constrain the SDF inference with the multi-view consistency. To this end, we dynamically align 3D Gaussians on the zero-level set of the neural SDF, and then render the aligned 3D Gaussians through the differentiable rasterization. Meanwhile, we update the neural SDF by pulling neighboring space to the pulled 3D Gaussians, which progressively refine the signed distance field near the surface. With both differentiable pulling and splatting, we jointly optimize 3D Gaussians and the neural SDF with both RGB and geometry constraints, which recovers more accurate, smooth, and complete surfaces with more geometry details. Our numerical and visual comparisons show our superiority over the state-of-the-art results on the widely used benchmarks.

Representing 3D scenes from multiview images remains a core challenge in computer vision and graphics, requiring both reliable rendering and reconstruction, which often conflicts due to the mismatched prioritization of image quality over precise underlying scene geometry. Although both neural implicit surfaces and explicit Gaussian primitives have advanced with neural rendering techniques, current methods impose strict constraints on density fields or primitive shapes, which enhances the affinity for geometric reconstruction at the sacrifice of rendering quality. To address this dilemma, we introduce GSDF, a dual-branch architecture combining 3D Gaussian Splatting (3DGS) and neural Signed Distance Fields (SDF). Our approach leverages mutual guidance and joint supervision during the training process to mutually enhance reconstruction and rendering. Specifically, our method guides the Gaussian primitives to locate near potential surfaces and accelerates the SDF convergence. This implicit mutual guidance ensures robustness and accuracy in both synthetic and real-world scenarios. Experimental results demonstrate that our method boosts the SDF optimization process to reconstruct more detailed geometry, while reducing floaters and blurry edge artifacts in rendering by aligning Gaussian primitives with the underlying geometry.

Recently, 3D Gaussian Splatting (3DGS) has become popular in reconstructing dense 3D representations of appearance and geometry. However, the learning pipeline in 3DGS inherently lacks the ability to quantify uncertainty, which is an important factor in applications like robotics mapping and navigation. In this paper, we propose an uncertainty estimation method built upon the Bayesian inference framework. Specifically, we propose a method to build variational multi-scale 3D Gaussians, where we leverage explicit scale information in 3DGS parameters to construct diversified parameter space samples. We develop an offset table technique to draw local multi-scale samples efficiently by offsetting selected attributes and sharing other base attributes. Then, the offset table is learned by variational inference with multi-scale prior. The learned offset posterior can quantify the uncertainty of each individual Gaussian component, and be used in the forward pass to infer the predictive uncertainty. Extensive experimental results on various benchmark datasets show that the proposed method provides well-aligned calibration performance on estimated uncertainty and better rendering quality compared with the previous methods that enable uncertainty quantification with view synthesis. Besides, by leveraging the model parameter uncertainty estimated by our method, we can remove noisy Gaussians automatically, thereby obtaining a high-fidelity part of the reconstructed scene, which is of great help in improving the visual quality.

The recent advancements in 3D Gaussian splatting (3D-GS) have not only facilitated real-time rendering through modern GPU rasterization pipelines but have also attained state-of-the-art rendering quality. Nevertheless, despite its exceptional rendering quality and performance on standard datasets, 3D-GS frequently encounters difficulties in accurately modeling specular and anisotropic components. This issue stems from the limited ability of spherical harmonics (SH) to represent high-frequency information. To overcome this challenge, we introduce Spec-Gaussian, an approach that utilizes an anisotropic spherical Gaussian (ASG) appearance field instead of SH for modeling the view-dependent appearance of each 3D Gaussian. Additionally, we have developed a coarse-to-fine training strategy to improve learning efficiency and eliminate floaters caused by overfitting in real-world scenes. Our experimental results demonstrate that our method surpasses existing approaches in terms of rendering quality. Thanks to ASG, we have significantly improved the ability of 3D-GS to model scenes with specular and anisotropic components without increasing the number of 3D Gaussians. This improvement extends the applicability of 3D GS to handle intricate scenarios with specular and anisotropic surfaces.

3D Hand reconstruction from a single RGB image is challenging due to the articulated motion, self-occlusion, and interaction with objects. Existing SOTA methods employ attention-based transformers to learn the 3D hand pose and shape, yet they do not fully achieve robust and accurate performance, primarily due to inefficiently modeling spatial relations between joints. To address this problem, we propose a novel graph-guided Mamba framework, named Hamba, which bridges graph learning and state space modeling. Our core idea is to reformulate Mamba's scanning into graph-guided bidirectional scanning for 3D reconstruction using a few effective tokens. This enables us to efficiently learn the spatial relationships between joints for improving reconstruction performance. Specifically, we design a Graph-guided State Space (GSS) block that learns the graph-structured relations and spatial sequences of joints and uses 88.5\% fewer tokens than attention-based methods. Additionally, we integrate the state space features and the global features using a fusion module. By utilizing the GSS block and the fusion module, Hamba effectively leverages the graph-guided state space features and jointly considers global and local features to improve performance. Experiments on several benchmarks and in-the-wild tests demonstrate that Hamba significantly outperforms existing SOTAs, achieving the PA-MPVPE of 5.3mm and F@15mm of 0.992 on FreiHAND. At the time of this paper's acceptance, Hamba holds the top position, Rank 1, in two competition leaderboards on 3D hand reconstruction.

Diffusion models have demonstrated their effectiveness in addressing the inherent uncertainty and indeterminacy in monocular 3D human pose estimation (HPE). Despite their strengths, the need for large search spaces and the corresponding demand for substantial training data make these models prone to generating biomechanically unrealistic poses. This challenge is particularly noticeable in occlusion scenarios, where the complexity of inferring 3D structures from 2D images intensifies. In response to these limitations, we introduce the **Di**screte **Di**ffusion **Pose** (**$\text{Di}^2\text{Pose}$**), a novel framework designed for occluded 3D HPE that capitalizes on the benefits of a discrete diffusion model. Specifically, **$\text{Di}^2\text{Pose}$** employs a two-stage process: it first converts 3D poses into a discrete representation through a pose quantization step, which is subsequently modeled in latent space through a discrete diffusion process. This methodological innovation restrictively confines the search space towards physically viable configurations and enhances the model’s capability to comprehend how occlusions affect human pose within the latent space. Extensive evaluations conducted on various benchmarks (e.g., Human3.6M, 3DPW, and 3DPW-Occ) have demonstrated its effectiveness.

Three-dimensional (3D) understanding of objects and scenes play a key role in humans' ability to interact with the world and has been an active area of research in computer vision, graphics, and robotics. Large scale synthetic and object-centric 3D datasets have shown to be effective in training models that have 3D understanding of objects. However, applying a similar approach to real-world objects and scenes is difficult due to a lack of large-scale data. Videos are a potential source for real-world 3D data, but finding diverse yet corresponding views of the same content have shown to be difficult at scale. Furthermore, standard videos come with fixed viewpoints, determined at the time of capture. This restricts the ability to access scenes from a variety of more diverse and potentially useful perspectives. We argue that large scale ODIN videos can address these limitations to provide scalable corresponding frames from diverse views. In this paper we introduce 360-1M, a 360° video dataset consisting of 1 million videos, and a process for efficiently finding corresponding frames from diverse viewpoints at scale. We train our diffusion-based model, ODIN, on 360-1M. Empowered by the largest real-world, multi-view dataset to date, ODIN is able to freely generate novel views of real-world scenes. Unlike previous methods, ODIN can move the camera through the environment, enabling the model to infer the geometry and layout of the scene. Additionally, we show improved performance on standard novel view synthesis and 3D reconstruction benchmarks.

Occupancy prediction has increasingly garnered attention in recent years for its fine-grained understanding of 3D scenes. Traditional approaches typically rely on dense, regular grid representations, which often leads to excessive computational demands and a loss of spatial details for small objects. This paper introduces OctreeOcc, an innovative 3D occupancy prediction framework that leverages the octree representation to adaptively capture valuable information in 3D, offering variable granularity to accommodate object shapes and semantic regions of varying sizes and complexities. In particular, we incorporate image semantic information to improve the accuracy of initial octree structures and design an effective rectification mechanism to refine the octree structure iteratively. Our extensive evaluations show that OctreeOcc not only surpasses state-of-the-art methods in occupancy prediction, but also achieves a 15%-24% reduction in computational overhead compared to dense-grid-based methods.

Novel view synthesis from sparse inputs is a vital yet challenging task in 3D computer vision. Previous methods explore 3D Gaussian Splatting with neural priors (e.g. depth priors) as an additional supervision, demonstrating promising quality and efficiency compared to the NeRF based methods. However, the neural priors from 2D pretrained models are often noisy and blurry, which struggle to precisely guide the learning of radiance fields. In this paper, We propose a novel method for synthesizing novel views from sparse views with Gaussian Splatting that does not require external prior as supervision. Our key idea lies in exploring the self-supervisions inherent in the binocular stereo consistency between each pair of binocular images constructed with disparity-guided image warping. To this end, we additionally introduce a Gaussian opacity constraint which regularizes the Gaussian locations and avoids Gaussian redundancy for improving the robustness and efficiency of inferring 3D Gaussians from sparse views. Extensive experiments on the LLFF, DTU, and Blender datasets demonstrate that our method significantly outperforms the state-of-the-art methods.

Determining the 3D orientations of an object in an image, known as single-image pose estimation, is a crucial task in 3D vision applications. Existing methods typically learn 3D rotations parametrized in the spatial domain using Euler angles or quaternions, but these representations often introduce discontinuities and singularities. SO(3)-equivariant networks enable the structured capture of pose patterns with data-efficient learning, but the parametrizations in spatial domain are incompatible with their architecture, particularly spherical CNNs, which operate in the frequency domain to enhance computational efficiency. To overcome these issues, we propose a frequency-domain approach that directly predicts Wigner-D coefficients for 3D rotation regression, aligning with the operations of spherical CNNs. Our SO(3)-equivariant pose harmonics predictor overcomes the limitations of spatial parameterizations, ensuring consistent pose estimation under arbitrary rotations. Trained with a frequency-domain regression loss, our method achieves state-of-the-art results on benchmarks such as ModelNet10-SO(3) and PASCAL3D+, with significant improvements in accuracy, robustness, and data efficiency.

The current trend in computer vision is to utilize one universal model to address all various tasks. Achieving such a universal model inevitably requires incorporating multi-domain data for joint training to learn across multiple problem scenarios. In point cloud based 3D object detection, however, such multi-domain joint training is highly challenging, because large domain gaps among point clouds from different datasets lead to the severe domain-interference problem. In this paper, we propose OneDet3D, a universal one-for-all model that addresses 3D detection across different domains, including diverse indoor and outdoor scenes, within the same framework and only one set of parameters. We propose the domain-aware partitioning in scatter and context, guided by a routing mechanism, to address the data interference issue, and further incorporate the text modality for a language-guided classification to unify the multi-dataset label spaces and mitigate the category interference issue. The fully sparse structure and anchor-free head further accommodate point clouds with significant scale disparities. Extensive experiments demonstrate the strong universal ability of OneDet3D to utilize only one trained model for addressing almost all 3D object detection tasks (Fig. 1). We will open-source the code for future research and applications.

Scene flow estimation is an essential ingredient for a variety of real-world applications, especially for autonomous agents, such as self-driving cars and robots. While recent scene flow estimation approaches achieve reasonable accuracy, their applicability to real-world systems additionally benefits from a reliability measure. Aiming at improving accuracy while additionally providing an estimate for uncertainty, we propose DiffSF that combines transformer-based scene flow estimation with denoising diffusion models. In the diffusion process, the ground truth scene flow vector field is gradually perturbed by adding Gaussian noise. In the reverse process, starting from randomly sampled Gaussian noise, the scene flow vector field prediction is recovered by conditioning on a source and a target point cloud. We show that the diffusion process greatly increases the robustness of predictions compared to prior approaches resulting in state-of-the-art performance on standard scene flow estimation benchmarks. Moreover, by sampling multiple times with different initial states, the denoising process predicts multiple hypotheses, which enables measuring the output uncertainty, allowing our approach to detect a majority of the inaccurate predictions. The code is available at https://github.com/ZhangYushan3/DiffSF.

Although the neural radiance field (NeRF) exhibits high-fidelity visualization on the rendering task, it still suffers from rendering defects, especially in complex scenes. In this paper, we delve into the reason for the unsatisfactory performance and conjecture that it comes from interference in the training process. Due to occlusions in complex scenes, a 3D point may be invisible to some rays. On such a point, training with those rays that do not contain valid information about the point might interfere with the NeRF training. Based on the above intuition, we decouple the training process of NeRF in the ray dimension softly and propose a Ray-decoupled Training Framework for neural rendering (Rad-NeRF). Specifically, we construct an ensemble of sub-NeRFs and train a soft gate module to assign the gating scores to these sub-NeRFs based on specific rays. The gate module is jointly optimized with the sub-NeRF ensemble to learn the preference of sub-NeRFs for different rays automatically. Furthermore, we introduce depth-based mutual learning to enhance the rendering consistency among multiple sub-NeRFs and mitigate the depth ambiguity. Experiments on five datasets demonstrate that Rad-NeRF can enhance the rendering performance across a wide range of scene types compared with existing single-NeRF and multi-NeRF methods. With only 0.2% extra parameters, Rad-NeRF improves rendering performance by up to 1.5dB. Code is available at https://github.com/thu-nics/Rad-NeRF.

Empowering 3D Gaussian Splatting with generalization ability is appealing. However, existing generalizable 3D Gaussian Splatting methods are largely confined to narrow-range interpolation between stereo images due to their heavy backbones, thus lacking the ability to accurately localize 3D Gaussian and support free-view synthesis across wide view range. In this paper, we present a novel framework FreeSplat that is capable of reconstructing geometrically consistent 3D scenes from long sequence input towards free-view synthesis.Specifically, we firstly introduce Low-cost Cross-View Aggregation achieved by constructing adaptive cost volumes among nearby views and aggregating features using a multi-scale structure. Subsequently, we present the Pixel-wise Triplet Fusion to eliminate redundancy of 3D Gaussians in overlapping view regions and to aggregate features observed across multiple views. Additionally, we propose a simple but effective free-view training strategy that ensures robust view synthesis across broader view range regardless of the number of views. Our empirical results demonstrate state-of-the-art novel view synthesis peformances in both novel view rendered color maps quality and depth maps accuracy across different numbers of input views. We also show that FreeSplat performs inference more efficiently and can effectively reduce redundant Gaussians, offering the possibility of feed-forward large scene reconstruction without depth priors. Our code will be made open-source upon paper acceptance.

In this paper, we introduce the \textbf{USCILab3D dataset}, a large-scale, annotated outdoor dataset designed for versatile applications across multiple domains, including computer vision, robotics, and machine learning. The dataset was acquired using a mobile robot equipped with 5 cameras and a 32-beam, $360^{\circ}$ scanning LIDAR. The robot was teleoperated, over the course of a year and under a variety of weather and lighting conditions, through a rich variety of paths within the USC campus (229 acres = $\sim 92.7$ hectares). The raw data was annotated using state-of-the-art large foundation models, and processed to provide multi-view imagery, 3D reconstructions, semantically-annotated images and point clouds (267 semantic categories), and text descriptions of images and objects within. The dataset also offers a diverse array of complex analyses using pose-stamping and trajectory data. In sum, the dataset offers 1.4M point clouds and 10M images ($\sim 6$TB of data). Despite covering a narrower geographical scope compared to a whole-city dataset, our dataset prioritizes intricate intersections along with denser multi-view scene images and semantic point clouds, enabling more precise 3D labelling and facilitating a broader spectrum of 3D vision tasks. For data, code and more details, please visit our website.

Digitally reconstructed radiographs (DRRs) are simulated 2D X-ray images generated from 3D CT volumes, widely used in preoperative settings but limited in intraoperative applications due to computational bottlenecks. Physics-based Monte Carlo simulations provide accurate representations but are extremely computationally intensity. Analytical DRR renderers are much more efficient, but at the price of ignoring anisotropic X-ray image formation phenomena such as Compton scattering. We propose a novel approach that balances realistic physics-inspired X-ray simulation with efficient, differentiable DRR generation using 3D Gaussian splatting (3DGS). Our direction-disentangled 3DGS (DDGS) method decomposes the radiosity contribution into isotropic and direction-dependent components, able to approximate complex anisotropic interactions without complex runtime simulations. Additionally, we adapt the 3DGS initialization to account for tomography data properties, enhancing accuracy and efficiency. Our method outperforms state-of-the-art techniques in image accuracy and inference speed, demonstrating its potential for intraoperative applications and inverse problems like pose registration.

The field of novel view synthesis from images has seen rapid advancements with the introduction of Neural Radiance Fields (NeRF) and more recently with 3D Gaussian Splatting. Gaussian Splatting became widely adopted due to its efficiency and ability to render novel views accurately. While Gaussian Splatting performs well when a sufficient amount of training images are available, its unstructured explicit representation tends to overfit in scenarios with sparse input images, resulting in poor rendering performance. To address this, we present a 3D Gaussian-based novel view synthesis method using sparse input images that can accurately render the scene from the viewpoints not covered by the training images. We propose a multi-stage training scheme with matching-based consistency constraints imposed on the novel views without relying on pre-trained depth estimation or diffusion models. This is achieved by using the matches of the available training images to supervise the generation of the novel views sampled between the training frames with color, geometry, and semantic losses. In addition, we introduce a locality preserving regularization for 3D Gaussians which removes rendering artifacts by preserving the local color structure of the scene. Evaluation on synthetic and real-world datasets demonstrates competitive or superior performance of our method in few-shot novel view synthesis compared to existing state-of-the-art methods.