NeurIPS 2024 论文列表

We introduce a radiance representation that is both structured and fully explicit and thus greatly facilitates 3D generative modeling. Existing radiance representations either require an implicit feature decoder, which significantly degrades the modeling power of the representation, or are spatially unstructured, making them difficult to integrate with mainstream 3D diffusion methods. We derive GaussianCube by first using a novel densification-constrained Gaussian fitting algorithm, which yields high-accuracy fitting using a fixed number of free Gaussians, and then rearranging these Gaussians into a predefined voxel grid via Optimal Transport. Since GaussianCube is a structured grid representation, it allows us to use standard 3D U-Net as our backbone in diffusion modeling without elaborate designs. More importantly, the high-accuracy fitting of the Gaussians allows us to achieve a high-quality representation with orders of magnitude fewer parameters than previous structured representations for comparable quality, ranging from one to two orders of magnitude. The compactness of GaussianCube greatly eases the difficulty of 3D generative modeling. Extensive experiments conducted on unconditional and class-conditioned object generation, digital avatar creation, and text-to-3D synthesis all show that our model achieves state-of-the-art generation results both qualitatively and quantitatively, underscoring the potential of GaussianCube as a highly accurate and versatile radiance representation for 3D generative modeling.

Vision language models (VLMs) have achieved impressive progress in diverse applications, becoming a prevalent research direction. In this paper, we build FIRE, a feedback-refinement dataset, consisting of 1.1M multi-turn conversations that are derived from 27 source datasets, empowering VLMs to spontaneously refine their responses based on user feedback across diverse tasks. To scale up the data collection, FIRE is collected in two components: FIRE-100K and FIRE-1M, where FIRE-100K is generated by GPT-4V, and FIRE-1M is freely generated via models trained on FIRE-100K. Then, we build FIRE-Bench, a benchmark to comprehensively evaluate the feedback-refining capability of VLMs, which contains 11K feedback-refinement conversations as the test data, two evaluation settings, and a model to provide feedback for VLMs. We develop the FIRE-LLaVA model by fine-tuning LLaVA on FIRE-100K and FIRE-1M, which shows remarkable feedback-refining capability on FIRE-Bench and outperforms untrained VLMs by 50%, making more efficient user-agent interactions and underscoring the significance of the FIRE dataset.

Omnidirectional (or 360-degree) images are increasingly being used for 3D applications since they allow the rendering of an entire scene with a single image. Existing works based on neural radiance fields demonstrate successful 3D reconstruction quality on egocentric videos, yet they suffer from long training and rendering times. Recently, 3D Gaussian splatting has gained attention for its fast optimization and real-time rendering. However, directly using a perspective rasterizer to omnidirectional images results in severe distortion due to the different optical properties between the two image domains. In this work, we present ODGS, a novel rasterization pipeline for omnidirectional images with geometric interpretation. For each Gaussian, we define a tangent plane that touches the unit sphere and is perpendicular to the ray headed toward the Gaussian center. We then leverage a perspective camera rasterizer to project the Gaussian onto the corresponding tangent plane. The projected Gaussians are transformed and combined into the omnidirectional image, finalizing the omnidirectional rasterization process. This interpretation reveals the implicit assumptions within the proposed pipeline, which we verify through mathematical proofs. The entire rasterization process is parallelized using CUDA, achieving optimization and rendering speeds 100 times faster than NeRF-based methods. Our comprehensive experiments highlight the superiority of ODGS by delivering the best reconstruction and perceptual quality across various datasets. Additionally, results on roaming datasets demonstrate that ODGS effectively restores fine details, even when reconstructing large 3D scenes. The source code is available on our project page (https://github.com/esw0116/ODGS).

Neural Radiance Fields (NeRFs) have shown remarkable performances in producing novel-view images from high-quality scene images. However, hand-held low-light photography challenges NeRFs as the captured images may simultaneously suffer from low visibility, noise, and camera shakes. While existing NeRF methods may handle either low light or motion, directly combining them or incorporating additional image-based enhancement methods does not work as these degradation factors are highly coupled. We observe that noise in low-light images is always sharp regardless of camera shakes, which implies an implicit order of these degradation factors within the image formation process. This inspires us to explore such an order to decouple and remove these degradation factors while training the NeRF. To this end, we propose in this paper a novel model, named LuSh-NeRF, which can reconstruct a clean and sharp NeRF from a group of hand-held low-light images. The key idea of LuSh-NeRF is to sequentially model noise and blur in the images via multi-view feature consistency and frequency information of NeRF, respectively. Specifically, LuSh-NeRF includes a novel Scene-Noise Decomposition (SND) module for decoupling the noise from the scene representation and a novel Camera Trajectory Prediction (CTP) module for the estimation of camera motions based on low-frequency scene information. To facilitate training and evaluations, we construct a new dataset containing both synthetic and real images. Experiments show that LuSh-NeRF outperforms existing approaches. Our code and dataset can be found here: https://github.com/quzefan/LuSh-NeRF.

Seams, distortions, wasted UV space, vertex-duplication, and varying resolution over the surface are the most prominent issues of the standard UV-based texturing of meshes. These issues are particularly acute when automatic UV-unwrapping techniques are used. For this reason, instead of generating textures in automatically generated UV-planes like most state-of-the-art methods, we propose to represent textures as coloured point-clouds whose colours are generated by a denoising diffusion probabilistic model constrained to operate on the surface of 3D objects. Our sampling and resolution agnostic generative model heavily relies on heat diffusion over the surface of the meshes for spatial communication between points. To enable processing of arbitrarily sampled point-cloud textures and ensure long-distance texture consistency we introduce a fast re-sampling of the mesh spectral properties used during the heat diffusion and introduce a novel heat-diffusion-based self-attention mechanism. Our code and pre-trained models are available at github.com/simofoti/UV3-TeD.

Existing human rendering methods require every part of the human to be fully visible throughout the input video. However, this assumption does not hold in real-life settings where obstructions are common, resulting in only partial visibility of the human. Considering this, we present OccFusion, an approach that utilizes efficient 3D Gaussian splatting supervised by pretrained 2D diffusion models for efficient and high-fidelity human rendering. We propose a pipeline consisting of three stages. In the Initialization stage, complete human masks are generated from partial visibility masks. In the Optimization stage, 3D human Gaussians are optimized with additional supervisions by Score-Distillation Sampling (SDS) to create a complete geometry of the human. Finally, in the Refinement stage, in-context inpainting is designed to further improve rendering quality on the less observed human body parts. We evaluate OccFusion on ZJU-MoCap and challenging OcMotion sequences and found that it achieves state-of-the-art performance in the rendering of occluded humans.

The creation of complex 3D scenes tailored to user specifications has been a tedious and challenging task with traditional 3D modeling tools. Although some pioneering methods have achieved automatic text-to-3D generation, they are generally limited to small-scale scenes with restricted control over the shape and texture. We introduce SceneCraft, a novel method for generating detailed indoor scenes that adhere to textual descriptions and spatial layout preferences provided by users. Central to our method is a rendering-based technique, which converts 3D semantic layouts into multi-view 2D proxy maps. Furthermore, we design a semantic and depth conditioned diffusion model to generate multi-view images, which are used to learn a neural radiance field (NeRF) as the final scene representation. Without the constraints of panorama image generation, we surpass previous methods in supporting complicated indoor space generation beyond a single room, even as complicated as a whole multi-bedroom apartment with irregular shapes and layouts. Through experimental analysis, we demonstrate that our method significantly outperforms existing approaches in complex indoor scene generation with diverse textures, consistent geometry, and realistic visual quality.

In this paper, we propose to create animatable avatars for interacting hands with 3D Gaussian Splatting (GS) and single-image inputs. Existing GS-based methods designed for single subjects often yield unsatisfactory results due to limited input views, various hand poses, and occlusions. To address these challenges, we introduce a novel two-stage interaction-aware GS framework that exploits cross-subject hand priors and refines 3D Gaussians in interacting areas. Particularly, to handle hand variations, we disentangle the 3D presentation of hands into optimization-based identity maps and learning-based latent geometric features and neural texture maps. Learning-based features are captured by trained networks to provide reliable priors for poses, shapes, and textures, while optimization-based identity maps enable efficient one-shot fitting of out-of-distribution hands. Furthermore, we devise an interaction-aware attention module and a self-adaptive Gaussian refinement module. These modules enhance image rendering quality in areas with intra- and inter-hand interactions, overcoming the limitations of existing GS-based methods. Our proposed method is validated via extensive experiments on the large-scale InterHand2.6M dataset, and it significantly improves the state-of-the-art performance in image quality. Code and models will be released upon acceptance.

Correspondences in point cloud registration are prone to outliers, significantly reducing registration accuracy and highlighting the need for precise inlier identification. In this paper, we propose a robust inlier identification algorithm for point cloud registration by reformulating the conventional registration problem as an alignment error $\ell_0$-minimization problem. The $\ell_0$-minimization problem is formulated for each local set, where those local sets are built on a compatibility graph of input correspondences. To resolve the $\ell_0$-minimization, we develop a novel two-stage decoupling strategy, which first decouples the alignment error into a rotation fitting error and a translation fitting error. Second, null-space matrices are employed to decouple inlier identification from the estimation of rotation and translation respectively, thereby applying Bayesian theory to $\ell_0$-minimization problems and solving for fitting errors. Correspondences with the smallest errors are identified as inliers to generate a transformation hypothesis for each local set. The best hypothesis is selected to perform registration. We demonstrate that the proposed inlier identification algorithm is robust under high outlier ratios and noise through experiments. Extensive results on the KITTI, 3DMatch, and 3DLoMatch datasets demonstrate that our method achieves state-of-the-art performance compared to both traditional and learning-based methods in various indoor and outdoor scenes.

We introduce MVSplat360, a feed-forward approach for 360° novel view synthesis (NVS) of diverse real-world scenes, using only sparse observations. This setting is inherently ill-posed due to minimal overlap among input views and insufficient visual information provided, making it challenging for conventional methods to achieve high-quality results. Our MVSplat360 addresses this by effectively combining geometry-aware 3D reconstruction with temporally consistent video generation. Specifically, it refactors a feed-forward 3D Gaussian Splatting (3DGS) model to render features directly into the latent space of a pre-trained Stable Video Diffusion (SVD) model, where these features then act as pose and visual cues to guide the denoising process and produce photorealistic 3D-consistent views. Our model is end-to-end trainable and supports rendering arbitrary views with as few as 5 sparse input views. To evaluate MVSplat360's performance, we introduce a new benchmark using the challenging DL3DV-10K dataset, where MVSplat360 achieves superior visual quality compared to state-of-the-art methods on wide-sweeping or even 360° NVS tasks. Experiments on the existing benchmark RealEstate10K also confirm the effectiveness of our model. Readers are highly recommended to view the video results at [donydchen.github.io/mvsplat360](https://donydchen.github.io/mvsplat360).

To train a deblurring network, an appropriate dataset with paired blurry and sharp images is essential. Existing datasets collect blurry images either synthetically by aggregating consecutive sharp frames or using sophisticated camera systems to capture real blur. However, these methods offer limited diversity in blur types (blur trajectories) or require extensive human effort to reconstruct large-scale datasets, failing to fully reflect real-world blur scenarios. To address this, we propose GS-Blur, a dataset of synthesized realistic blurry images created using a novel approach. To this end, we first reconstruct 3D scenes from multi-view images using 3D Gaussian Splatting~(3DGS), then render blurry images by moving the camera view along the randomly generated motion trajectories. By adopting various camera trajectories in reconstructing our GS-Blur, our dataset contains realistic and diverse types of blur, offering a large-scale dataset that generalizes well to real-world blur. Using GS-Blur with various deblurring methods, we demonstrate its ability to generalize effectively compared to previous synthetic or real blur datasets, showing significant improvements in deblurring performance. We will publicly release our dataset.

While humans effortlessly discern intrinsic dynamics and adapt to new scenarios, modern AI systems often struggle. Current methods for visual grounding of dynamics either use pure neural-network-based simulators (black box), which may violate physical laws, or traditional physical simulators (white box), which rely on expert-defined equations that may not fully capture actual dynamics. We propose the Neural Material Adaptor (NeuMA), which integrates existing physical laws with learned corrections, facilitating accurate learning of actual dynamics while maintaining the generalizability and interpretability of physical priors. Additionally, we propose Particle-GS, a particle-driven 3D Gaussian Splatting variant that bridges simulation and observed images, allowing back-propagate image gradients to optimize the simulator. Comprehensive experiments on various dynamics in terms of grounded particle accuracy, dynamic rendering quality, and generalization ability demonstrate that NeuMA can accurately capture intrinsic dynamics. Project Page: https://xjay18.github.io/projects/neuma.html.

Recent 3D large reconstruction models (LRMs) can generate high-quality 3D content in sub-seconds by integrating multi-view diffusion models with scalable multi-view reconstructors. Current works further leverage 3D Gaussian Splatting as 3D representation for improved visual quality and rendering efficiency. However, we observe that existing Gaussian reconstruction models often suffer from multi-view inconsistency and blurred textures. We attribute this to the compromise of multi-view information propagation in favor of adopting powerful yet computationally intensive architectures (\eg, Transformers). To address this issue, we introduce MVGamba, a general and lightweight Gaussian reconstruction model featuring a multi-view Gaussian reconstructor based on the RNN-like State Space Model (SSM). Our Gaussian reconstructor propagates causal context containing multi-view information for cross-view self-refinement while generating a long sequence of Gaussians for fine-detail modeling with linear complexity. With off-the-shelf multi-view diffusion models integrated, MVGamba unifies 3D generation tasks from a single image, sparse images, or text prompts. Extensive experiments demonstrate that MVGamba outperforms state-of-the-art baselines in all 3D content generation scenarios with approximately only $0.1\times$ of the model size. The codes are available at \url{https://github.com/SkyworkAI/MVGamba}.

Within the realm of Computer-Aided Design (CAD), Boundary-Representation (B-Rep) is the standard option for modeling shapes. We present SpelsNet, a neural architecture for the segmentation of 3D point clouds into surface primitive elements under topological supervision of its B-Rep graph structure. We also propose a point-to-BRep adjacency representation that allows for adapting conventional Linear Algebraic Representation of B-Rep graph structure to the point cloud domain. Thanks to this representation, SpelsNet learns from both spatial and topological domains to enable accurate and topologically consistent surface primitive element segmentation. In particular, SpelsNet is composed of two main components; (1) a supervised 3D spatial segmentation head that outputs B-Rep element types and memberships; (2) a graph-based head that leverages the proposed topological supervision. To enable the learning of SpelsNet with the proposed point-to-BRep adjacency supervision, we extend two existing CAD datasets with the required annotations, and conduct a thorough experimental validation on them. The obtained results showcase the efficacy of SpelsNet and its topological supervision compared to a set of baselines and state-of-the-art approaches.

Existing open-set learning methods consider only the single-layer labels of objects and strictly assume no overlap between the training and testing sets, leading to contradictory optimization for superposed categories. In this paper, we introduce a more practical Semi-Open Environment setting for open-set 3D object retrieval with hierarchical labels, in which the training and testing set share a partial label space for coarse categories but are completely disjoint from fine categories. We propose the Hypergraph-Based Hierarchical Equilibrium Representation (HERT) framework for this task. Specifically, we propose the Hierarchical Retrace Embedding (HRE) module to overcome the global disequilibrium of unseen categories by fully leveraging the multi-level category information. Besides, tackling the feature overlap and class confusion problem, we perform the Structured Equilibrium Tuning (SET) module to utilize more equilibrial correlations among objects and generalize to unseen categories, by constructing a superposed hypergraph based on the local coherent and global entangled correlations. Furthermore, we generate four semi-open 3DOR datasets with multi-level labels for benchmarking. Results demonstrate that the proposed method can effectively generate the hierarchical embeddings of 3D objects and generalize them towards semi-open environments.

We propose a novel unsupervised method to learn pose and part-segmentation of articulated objects with rigid parts. Given two observations of an object in different articulation states, our method learns the geometry and appearance of object parts by using an implicit model from the first observation, distills the part segmentation and articulation from the second observation while rendering the latter observation. Additionally, to tackle the complexities in the joint optimization of part segmentation and articulation, we propose a voxel grid based initialization strategy and a decoupled optimization procedure. Compared to the prior unsupervised work, our model obtains significantly better performance, generalizes to objects with multiple parts while it can be efficiently from few views for the latter observation.

Unsupervised 3D object detection methods have emerged to leverage vast amounts of data without requiring manual labels for training. Recent approaches rely on dynamic objects for learning to detect mobile objects but penalize the detections of static instances during training. Multiple rounds of (self) training are used to add detected static instances to the set of training targets; this procedure to improve performance is computationally expensive. To address this, we propose the method UNION. We use spatial clustering and self-supervised scene flow to obtain a set of static and dynamic object proposals from LiDAR. Subsequently, object proposals' visual appearances are encoded to distinguish static objects in the foreground and background by selecting static instances that are visually similar to dynamic objects. As a result, static and dynamic mobile objects are obtained together, and existing detectors can be trained with a single training. In addition, we extend 3D object discovery to detection by using object appearance-based cluster labels as pseudo-class labels for training object classification. We conduct extensive experiments on the nuScenes dataset and increase the state-of-the-art performance for unsupervised 3D object discovery, i.e. UNION more than doubles the average precision to 38.4. The code is available at github.com/TedLentsch/UNION.