Abstract:Autoregressive models have achieved remarkable success across various domains, yet their performance in 3D shape generation lags significantly behind that of diffusion models. In this paper, we introduce OctGPT, a novel multiscale autoregressive model for 3D shape generation that dramatically improves the efficiency and performance of prior 3D autoregressive approaches, while rivaling or surpassing state-of-the-art diffusion models. Our method employs a serialized octree representation to efficiently capture the hierarchical and spatial structures of 3D shapes. Coarse geometry is encoded via octree structures, while fine-grained details are represented by binary tokens generated using a vector quantized variational autoencoder (VQVAE), transforming 3D shapes into compact multiscale binary sequences suitable for autoregressive prediction. To address the computational challenges of handling long sequences, we incorporate octree-based transformers enhanced with 3D rotary positional encodings, scale-specific embeddings, and token-parallel generation schemes. These innovations reduce training time by 13 folds and generation time by 69 folds, enabling the efficient training of high-resolution 3D shapes, e.g.,$1024^3$, on just four NVIDIA 4090 GPUs only within days. OctGPT showcases exceptional versatility across various tasks, including text-, sketch-, and image-conditioned generation, as well as scene-level synthesis involving multiple objects. Extensive experiments demonstrate that OctGPT accelerates convergence and improves generation quality over prior autoregressive methods, offering a new paradigm for high-quality, scalable 3D content creation. Our code and trained models are available at https://github.com/octree-nn/octgpt.
Abstract:Accurately reconstructing continuous flow fields from sparse or indirect measurements remains an open challenge, as existing techniques often suffer from oversmoothing artifacts, reliance on heterogeneous architectures, and the computational burden of enforcing physics-informed losses in implicit neural representations (INRs). In this paper, we introduce a novel flow field reconstruction framework based on divergence-free kernels (DFKs), which inherently enforce incompressibility while capturing fine structures without relying on hierarchical or heterogeneous representations. Through qualitative analysis and quantitative ablation studies, we identify the matrix-valued radial basis functions derived from Wendland's $\mathcal{C}^4$ polynomial (DFKs-Wen4) as the optimal form of analytically divergence-free approximation for velocity fields, owing to their favorable numerical properties, including compact support, positive definiteness, and second-order differentiablility. Experiments across various reconstruction tasks, spanning data compression, inpainting, super-resolution, and time-continuous flow inference, has demonstrated that DFKs-Wen4 outperform INRs and other divergence-free representations in both reconstruction accuracy and computational efficiency while requiring the fewest trainable parameters.
Abstract:We consider the problem of adding dynamic rain effects to in-the-wild scenes in a physically-correct manner. Recent advances in scene modeling have made significant progress, with NeRF and 3DGS techniques emerging as powerful tools for reconstructing complex scenes. However, while effective for novel view synthesis, these methods typically struggle with challenging scene editing tasks, such as physics-based rain simulation. In contrast, traditional physics-based simulations can generate realistic rain effects, such as raindrops and splashes, but they often rely on skilled artists to carefully set up high-fidelity scenes. This process lacks flexibility and scalability, limiting its applicability to broader, open-world environments. In this work, we introduce RainyGS, a novel approach that leverages the strengths of both physics-based modeling and 3DGS to generate photorealistic, dynamic rain effects in open-world scenes with physical accuracy. At the core of our method is the integration of physically-based raindrop and shallow water simulation techniques within the fast 3DGS rendering framework, enabling realistic and efficient simulations of raindrop behavior, splashes, and reflections. Our method supports synthesizing rain effects at over 30 fps, offering users flexible control over rain intensity -- from light drizzles to heavy downpours. We demonstrate that RainyGS performs effectively for both real-world outdoor scenes and large-scale driving scenarios, delivering more photorealistic and physically-accurate rain effects compared to state-of-the-art methods. Project page can be found at https://pku-vcl-geometry.github.io/RainyGS/
Abstract:In this paper, we propose ProTracker, a novel framework for robust and accurate long-term dense tracking of arbitrary points in videos. The key idea of our method is incorporating probabilistic integration to refine multiple predictions from both optical flow and semantic features for robust short-term and long-term tracking. Specifically, we integrate optical flow estimations in a probabilistic manner, producing smooth and accurate trajectories by maximizing the likelihood of each prediction. To effectively re-localize challenging points that disappear and reappear due to occlusion, we further incorporate long-term feature correspondence into our flow predictions for continuous trajectory generation. Extensive experiments show that ProTracker achieves the state-of-the-art performance among unsupervised and self-supervised approaches, and even outperforms supervised methods on several benchmarks. Our code and model will be publicly available upon publication.
Abstract:In this paper, we introduce \textbf{SLAM3R}, a novel and effective monocular RGB SLAM system for real-time and high-quality dense 3D reconstruction. SLAM3R provides an end-to-end solution by seamlessly integrating local 3D reconstruction and global coordinate registration through feed-forward neural networks. Given an input video, the system first converts it into overlapping clips using a sliding window mechanism. Unlike traditional pose optimization-based methods, SLAM3R directly regresses 3D pointmaps from RGB images in each window and progressively aligns and deforms these local pointmaps to create a globally consistent scene reconstruction - all without explicitly solving any camera parameters. Experiments across datasets consistently show that SLAM3R achieves state-of-the-art reconstruction accuracy and completeness while maintaining real-time performance at 20+ FPS. Code and weights at: \url{https://github.com/PKU-VCL-3DV/SLAM3R}.
Abstract:We consider the problem of physically-based inverse rendering using 3D Gaussian Splatting (3DGS) representations. While recent 3DGS methods have achieved remarkable results in novel view synthesis (NVS), accurately capturing high-fidelity geometry, physically interpretable materials and lighting remains challenging, as it requires precise geometry modeling to provide accurate surface normals, along with physically-based rendering (PBR) techniques to ensure correct material and lighting disentanglement. Previous 3DGS methods resort to approximating surface normals, but often struggle with noisy local geometry, leading to inaccurate normal estimation and suboptimal material-lighting decomposition. In this paper, we introduce GeoSplatting, a novel hybrid representation that augments 3DGS with explicit geometric guidance and differentiable PBR equations. Specifically, we bridge isosurface and 3DGS together, where we first extract isosurface mesh from a scalar field, then convert it into 3DGS points and formulate PBR equations for them in a fully differentiable manner. In GeoSplatting, 3DGS is grounded on the mesh geometry, enabling precise surface normal modeling, which facilitates the use of PBR frameworks for material decomposition. This approach further maintains the efficiency and quality of NVS from 3DGS while ensuring accurate geometry from the isosurface. Comprehensive evaluations across diverse datasets demonstrate the superiority of GeoSplatting, consistently outperforming existing methods both quantitatively and qualitatively.
Abstract:2D irregular packing is a classic combinatorial optimization problem with various applications, such as material utilization and texture atlas generation. This NP-hard problem requires efficient algorithms to optimize space utilization. Conventional numerical methods suffer from slow convergence and high computational cost. Existing learning-based methods, such as the score-based diffusion model, also have limitations, such as no rotation support, frequent collisions, and poor adaptability to arbitrary boundaries, and slow inferring. The difficulty of learning from teacher packing is to capture the complex geometric relationships among packing examples, which include the spatial (position, orientation) relationships of objects, their geometric features, and container boundary conditions. Representing these relationships in latent space is challenging. We propose GFPack++, an attention-based gradient field learning approach that addresses this challenge. It consists of two pivotal strategies: \emph{attention-based geometry encoding} for effective feature encoding and \emph{attention-based relation encoding} for learning complex relationships. We investigate the utilization distribution between the teacher and inference data and design a weighting function to prioritize tighter teacher data during training, enhancing learning effectiveness. Our diffusion model supports continuous rotation and outperforms existing methods on various datasets. We achieve higher space utilization over several widely used baselines, one-order faster than the previous diffusion-based method, and promising generalization for arbitrary boundaries. We plan to release our source code and datasets to support further research in this direction.
Abstract:3D Gaussian Splatting-based techniques have recently advanced 3D scene reconstruction and novel view synthesis, achieving high-quality real-time rendering. However, these approaches are inherently limited by the underlying pinhole camera assumption in modeling the images and hence only work for All-in-Focus (AiF) sharp image inputs. This severely affects their applicability in real-world scenarios where images often exhibit defocus blur due to the limited depth-of-field (DOF) of imaging devices. Additionally, existing 3D Gaussian Splatting (3DGS) methods also do not support rendering of DOF effects. To address these challenges, we introduce DOF-GS that allows for rendering adjustable DOF effects, removing defocus blur as well as refocusing of 3D scenes, all from multi-view images degraded by defocus blur. To this end, we re-imagine the traditional Gaussian Splatting pipeline by employing a finite aperture camera model coupled with explicit, differentiable defocus rendering guided by the Circle-of-Confusion (CoC). The proposed framework provides for dynamic adjustment of DOF effects by changing the aperture and focal distance of the underlying camera model on-demand. It also enables rendering varying DOF effects of 3D scenes post-optimization, and generating AiF images from defocused training images. Furthermore, we devise a joint optimization strategy to further enhance details in the reconstructed scenes by jointly optimizing rendered defocused and AiF images. Our experimental results indicate that DOF-GS produces high-quality sharp all-in-focus renderings conditioned on inputs compromised by defocus blur, with the training process incurring only a modest increase in GPU memory consumption. We further demonstrate the applications of the proposed method for adjustable defocus rendering and refocusing of the 3D scene from input images degraded by defocus blur.
Abstract:In this work, we present Semantic Gesticulator, a novel framework designed to synthesize realistic gestures accompanying speech with strong semantic correspondence. Semantically meaningful gestures are crucial for effective non-verbal communication, but such gestures often fall within the long tail of the distribution of natural human motion. The sparsity of these movements makes it challenging for deep learning-based systems, trained on moderately sized datasets, to capture the relationship between the movements and the corresponding speech semantics. To address this challenge, we develop a generative retrieval framework based on a large language model. This framework efficiently retrieves suitable semantic gesture candidates from a motion library in response to the input speech. To construct this motion library, we summarize a comprehensive list of commonly used semantic gestures based on findings in linguistics, and we collect a high-quality motion dataset encompassing both body and hand movements. We also design a novel GPT-based model with strong generalization capabilities to audio, capable of generating high-quality gestures that match the rhythm of speech. Furthermore, we propose a semantic alignment mechanism to efficiently align the retrieved semantic gestures with the GPT's output, ensuring the naturalness of the final animation. Our system demonstrates robustness in generating gestures that are rhythmically coherent and semantically explicit, as evidenced by a comprehensive collection of examples. User studies confirm the quality and human-likeness of our results, and show that our system outperforms state-of-the-art systems in terms of semantic appropriateness by a clear margin.
Abstract:This letter introduces a novel framework for dense Visual Simultaneous Localization and Mapping (VSLAM) based on Gaussian Splatting. Recently Gaussian Splatting-based SLAM has yielded promising results, but rely on RGB-D input and is weak in tracking. To address these limitations, we uniquely integrates advanced sparse visual odometry with a dense Gaussian Splatting scene representation for the first time, thereby eliminating the dependency on depth maps typical of Gaussian Splatting-based SLAM systems and enhancing tracking robustness. Here, the sparse visual odometry tracks camera poses in RGB stream, while Gaussian Splatting handles map reconstruction. These components are interconnected through a Multi-View Stereo (MVS) depth estimation network. And we propose a depth smooth loss to reduce the negative effect of estimated depth maps. Furthermore, the consistency in scale between the sparse visual odometry and the dense Gaussian map is preserved by Sparse-Dense Adjustment Ring (SDAR). We have evaluated our system across various synthetic and real-world datasets. The accuracy of our pose estimation surpasses existing methods and achieves state-of-the-art performance. Additionally, it outperforms previous monocular methods in terms of novel view synthesis fidelity, matching the results of neural SLAM systems that utilize RGB-D input.