We introduce the PhaseTree, a novel hierarchical construction-tree representation for compactly modeling volumetric objects composed of multiple phases or materials across scales. An object is defined as a single construction tree that combines phase-aware primitives through composition and warping operators, yielding a unified multiphase signed distance representation that naturally supports complex topologies and interfaces between phases. The PhaseTree is compatible with standard signed distance field workflows: single-phase algorithms can be directly promoted to a PhaseTree, and conversely reduced without loss of information. As a result, our model integrates seamlessly with existing algorithms and rendering pipelines. We extend classical Sphere Tracing to robustly handle multiphase configurations and show that, despite the additional expressiveness, our implementation preserves the compactness and resolution independence of signed distance fields and incurs less than a 25% runtime overhead compared to single-phase Sphere Tracing.
论文检索
输入标题、作者或关键词，从 17,487 篇学术成果中精准定位
Recent progress has shown that video diffusion models (VDMs) can be repurposed to solve various multimodal graphics tasks. However, existing approaches predominantly train separate models for each specific problem setting. This practice locks models into fixed input-output mappings, and typically ignores the joint correlations across modalities. In this paper, we present UniVidX, a unified multimodal framework designed to leverage VDM priors to enable versatile video generation. Our goal is to (i) master diverse pixel-aligned tasks by formulating them as conditional generation problems within multimodal space, (ii) adapt to modality-specific distributions without compromising the backbone's native priors, and (iii) ensure cross-modal consistency during synthesis. Concretely, we propose three key designs: 1) Stochastic Condition Masking (SCM): by randomly partitioning modalities into clean conditions and noisy targets during training, we enable the model to learn omni-directional conditional generation rather than fixed mappings. 2) Decoupled Gated LoRA (DGL): we attach per-modality LoRAs and activate them when a modality serves as a generation target, thereby preserving the VDM's strong priors. 3) Cross-Modal Self-Attention (CMSA): we explicitly share keys/values across modalities while maintaining modality-specific queries, facilitating information exchange and inter-modal alignment. We validate our framework by instantiating it in two domains: 1) UniVid-Intrinsic for RGB videos and their intrinsic maps (albedo, irradiance, normal), and 2) UniVid-Alpha for blended RGB videos and their constituent RGBA layers. Experimental results demonstrate that both models achieve performance competitive with state-of-the-art methods across distinct tasks. Notably, they exhibit robust generalization capabilities in in-the-wild scenarios, even when trained on limited datasets of fewer than 1k videos.
Computing the Hausdorff distance between triangle meshes with guaranteed accuracy is a computationally intensive task. Conventional Branch-and-Bound (B&B) approaches are fundamentally ill-suited for massive parallelism. Their reliance on a global priority queue (PQ) for both best-first scheduling and global termination checks creates a serial bottleneck that prevents scalable performance. We introduce PQ-Free HD, a parallel B&B framework that eliminates this dependency by decoupling the algorithm's termination logic from its scheduling order. This is achieved by relaxing the culling criterion, thereby replacing the priority queue with a contention-free ring buffer, which transforms the execution model from a state-dependent serial search into a high-throughput, asynchronous batch-processing paradigm. The framework consists of four key components: (1) a parallel priority-queue-free B&B paradigm; (2) a hierarchical GPU execution architecture combining batched depth-first scheduling with fused collaborative kernels; (3) a geometrically robust seven-stage culling pipeline featuring novel tests for challenging geometries; and (4) a compact 29-byte procedural task descriptor that achieves an 83.9% memory reduction. Evaluations demonstrate substantial speedups: a median of 71.7× over the state-of-the-art CPU algorithm on general benchmarks, and exceeding 10,000× on challenging CAD models with dense planar structures. The throughput advantage scales super-linearly with problem complexity. We showcase practical value by building a strictly Hausdorff-distance-bounded mesh simplification tool entirely on the GPU. Our work provides a new method for high-throughput, tolerance-controllable B&B-based geometric queries on GPUs. Code and data are available at https://github.com/huzhihao2001/pqfree-hd.
Computing the directed Hausdorff distance between two triangle meshes is a fundamental operation in geometry processing and simulation. While existing certified branch-and-bound (B&B) methods are efficient for well-separated geometry, they can become prohibitively expensive on large models under tight tolerances and near-zero distance configurations where pruning is limited. We present a GPU-accelerated certified B&B algorithm that explicitly maintains enclosing lower and upper bounds on the directed Hausdorff distance and terminates once their normalized gap, measured with respect to the bounding-box diagonal of the source mesh, meets a user-prescribed tolerance. To map the inherently prioritized search to SIMT (single-instruction, multiple-thread) hardware, we replace priority queues and recursion with a sorted, double-buffered wavefront pipeline built from bulk-parallel worklists for bound evaluation, culling, subdivision, and compaction. To mitigate loose bounds on thin primitives while preserving predictable stream behavior, we introduce a fixed-cardinality adaptive subdivision scheme that selectively applies double longest-edge bisection. To remain robust in deep-refinement regimes, we add a resource-aware deferral mechanism that enforces a device-capacity invariant by prioritizing candidates likely to be culled while postponing expensive ones. Finally, we improve numerical robustness under FP32 (single precision) via triangle-local coordinate transforms and other conservative numerical safeguards, and enhance coherence by spatially ordering the active set and traversing the BVH (bounding volume hierarchy) in triangle packets. Under the same stopping tolerance, experiments on an NVIDIA RTX 5090 show that our GPU solver remains numerically consistent with the FP64 CPU baseline, with normalized cross-platform deviation below 0.01% in over 99.9% of cases. Our method achieves millisecond-scale runtimes capable of supporting interactive frame rates, even on models with millions of triangles. Across the comparison set, it delivers throughput speedups of 836× on the Thingi10K/TetWild benchmark (A → B) and 709× on the Thingi10K/Decimation benchmark. Code and data for this paper are available at https://github.com/fhp-transient/gpu-hausdorff.
Generalized winding numbers provide a robust measure of point insidedness for 3D surfaces—whether open, self-intersecting, or non-manifold—and are central to numerous geometry processing tasks. However, existing methods trade off between accuracy and computational efficiency, limiting their use in interactive and large-scale applications. We introduce a new formulation and algorithm for computing generalized winding numbers that is both fast and accurate to arbitrary precision, applicable to meshes and parametric surfaces. Our approach expresses the winding number as the sum of two intuitive geometric quantities: the signed number of ray-surface intersections and a boundary integral over the surface's projection onto the unit sphere. This insight leads to an efficient discretization that avoids expensive surface integrals and spherical arrangements. For meshes, our method achieves average speedups of 22X on a CPU compared to the fastest precise methods and 3X compared to the fastest approximation method, while maintaining full precision. On a GPU, for moderately complex meshes we reach a throughput of 109 queries per second, or 4K generalized winding number slices at 120 FPS (13X faster than a naïve GPU method). For parametric surfaces, our method is on average 5.6X faster than the state-of-the-art method, with the same precision. Our method naturally handles complex topologies and non-manifold inputs. We extensively validate its accuracy, robustness, and time performance. Our code is available at https://github.com/MartensCedric/antipodal.
The generalized winding number (GWN) is a scalar field that supports robust containment queries on curved geometry, including non-watertight, overlapping, and nested boundary representations. While queries can be easily parallelized over samples, direct evaluation on parametric curves and surfaces remains costly for large and complex models. Fast, state-of-the-art GWN approaches leverage a spatial index to approximate the GWN, typically coupled with a Taylor expansion which approximates the GWN contribution for far clusters of geometric primitives. However, such methods operate only on discrete inputs such as triangle meshes and point clouds, and would introduce containment errors near boundaries if applied to curved input. We extend support for fast GWN evaluation over arbitrary collections of NURBS curves in 2D and trimmed NURBS patches in 3D via a Bounding Volume Hierarchy that stores efficiently precomputed moment data in the hierarchy nodes. When querying the hierarchy, approximations for far clusters are used alongside direct evaluation for nearby NURBS primitives, achieving sub-linear complexity while preserving the geometric features in the vicinity of the query point. Central to our performance improvements is an adaptive subdivision strategy for NURBS primitives during a preprocessing phase, creating better spatial partitions while retaining the same accuracy for containment decisions as a direct evaluation. We demonstrate the performance and accuracy of our approach across a large collection of 2D and 3D datasets.
Eigenpair extractions are crucial for various applications in geometry processing and graphics. State of the Art libraries like ARPACK or Spectra rely on the implicitly restarted Lanczos iteration to extract eigenpairs efficiently. However for some large scale problems they lack convergence speed and robustness. In this paper we present a simple multigrid extension to accelerate the convergence and robustness of the implicitly restarted Lanczos method, and we demonstrate the efficiency of our method on a variety of problems commonly found in geometry processing and graphics.
Online 3D reconstruction from monocular image sequences is a challenging and ongoing research topic. 3D Gaussian Splatting (3DGS), leveraging its high-quality real-time rendering capability, empowers online 3D reconstruction to represent dense scenes with enhanced expressiveness, and thus holds great promise for a wide range of applications such as robotics and AR/VR. However, existing online 3DGS methods still suffer from some key challenges: fragile camera pose estimation due to the lack of global optimization, and low optimization efficiency in large-scale or long-sequence scenarios. To address these issues, we propose a robust and efficient online voxelized 3DGS reconstruction framework integrated with global Sim(3) optimization, which enables reliable camera tracking and efficient global loop closure for both camera poses and voxelized 3DGS. To accelerate the convergence of the voxelized 3DGS, we further introduce a color residual learning strategy, which not only boosts optimization speed but also enhances rendering quality. Extensive experiments on diverse indoor and outdoor datasets demonstrate that our method achieves state-of-the-art performance in both camera pose estimation accuracy and rendering quality, while retaining real-time efficiency. Additionally, we develop and deploy a real-world UAV-based active reconstruction system grounded on our proposed method, validating its robustness and generalizability for practical online 3D reconstruction tasks. Our code and data are available at https://github.com/TrickyGo/MoonSplat..
Video-guided 3D animation holds immense potential for content creation, offering intuitive and precise control over dynamic assets. However, practical deployment faces a critical yet frequently overlooked hurdle: the pose misalignment dilemma. In real-world scenarios, the initial pose of a user-provided static mesh rarely aligns with the starting frame of a reference video. Naively forcing a mesh to follow a mismatched trajectory inevitably leads to severe geometric distortion or animation failure. To address this, we present Rectified Dynamic Mesh (R-DMesh), a unified framework designed to generate high-fidelity 4D meshes that are “rectified” to align with video context. Unlike standard motion transfer approaches, our method introduces a novel VAE that explicitly disentangles the input into a conditional base mesh, relative motion trajectories, and a crucial rectification jump offset. This offset is learned to automatically transform the arbitrary pose of the input mesh to match the video’s initial state before animation begins. We process these components via a Triflow Attention mechanism, which leverages vertex-wise geometric features to modulate the three orthogonal flows, ensuring physical consistency and local rigidity during the rectification and animation process. For generation, we employ a Rectified Flow-based Diffusion Transformer conditioned on pre-trained video latents, effectively transferring rich spatio-temporal priors to the 3D domain. To support this task, we construct Video-RDMesh, a large-scale dataset of over 500k dynamic mesh sequences specifically curated to simulate pose misalignment. Extensive experiments demonstrate that R-DMesh not only solves the alignment problem but also enables robust downstream applications, including pose retargeting and holistic 4D generation. Code and pre-trained weights will be available at: https://github.com/Tencent-Hunyuan/R-DMesh.
Tone mapping evaluation is difficult because of the substantial differences in absolute luminance between high dynamic range (HDR) reference and tone-mapped standard dynamic range (SDR) test content. To address this challenge, we collected a new tone mapping evaluation dataset, focused on fundamental tone mapping operations, and combined it with several existing tone mapping quality assessment datasets. Rather than introducing new specialized metrics designed for tone-mapped content, we instead developed a set of techniques to adapt existing quality metrics for tone mapping quality assessment. Our approach models the photometric differences between HDR reference and SDR test displays for accurate metric predictions. The technique consists of two steps: first, a display model converts display-encoded content to photometric values; second, these values are re-encoded using a perceptual transfer function to map both HDR and tone-mapped images to the same display-encoded color space. We systematically evaluated both general-purpose image and video quality metrics with our adaptations and those specifically designed for tone mapping. With these adjustments, general-purpose metrics perform much better for tone mapping evaluation, consistently outperforming previously established specialized techniques. Additionally, we adapted the ColorVideoVDP metric to be sensitive to absolute luminance changes, resulting in ColorVideoVDP-tm, which shows greatly improved performance and accepts photometric values as input. These results highlight the robustness of our adaptation technique and provide an improved protocol to evaluate future tone mapping quality metrics. Our datasets, code, and supplementary results can be found at kenchen10.github.io/projects/tmometric/index.html.
Adaptive densification is the engine of 3D Gaussian Splatting (3DGS). However, when transposed to the optimization-based Generative Distillation paradigm, this reconstruction-native mechanism reveals fundamental limitations, resulting in inefficient representations cluttered with redundant primitives. We diagnose this failure as a Densification Dilemma stemming from the stochastic nature of generative guidance: the standard magnitude-based accumulation indiscriminately aggregates transient noise alongside geometric signals, making it difficult to strike a balance between over-densification and under-fitting. To resolve this, we introduce Context-Adaptive Moment Estimation (CAdam), a novel framework that reinterprets densification as a statistically grounded signal verification problem. CAdam leverages the first moment of gradients to exploit the interference principle—where stochastic fluctuations cancel out via destructive interference while consistent geometric drifts accumulate via constructive interference—effectively disentangling the underlying signal from the generative noise floor. This is further augmented by a quantile-based context awareness and an intrinsic Signal-to-Noise Ratio (SNR) gating mechanism, which ensure robust adaptation across optimization stages and enable the soft termination of densification. Extensive experiments across diverse objectives (SDS, ISM, VFDS) and strong generative 3DGS backbones show that CAdam reduces Gaussian count by 85%–97% relative to standard densification while preserving overall comparable perceptual quality. These results highlight signal-aware density control as a practical way to improve memory efficiency in optimization-based generative distillation.
Selection is a core operation in interactive image editing, enabling tasks such as composition or manipulation. To be practically useful, a user should be able to specify and disambiguate the desired selection region through either text- or click-based interactions, and the system should support selecting not only objects but also other criteria, such as materials. Material-based selection can be particularly valuable for tasks like re-texturing surfaces or consistently editing all instances of a specific material in a scene. However, existing vision–language-model (VLM) based selection methods are largely object-centric and typically support only a single interaction modality, limiting their applicability in real editing workflows. In this work, we thus present Mask Any Object And Material (MAOAM), a unified selection framework that enables precise object- and material-level selection across both text- and click-based interactions. MAOAM leverages a VLM with a segmentation head to produce pixel-accurate masks from user prompts: the VLM interprets the user’s selection intent — object- or material-level — and encodes visual entities, attributes, and spatial relations, while the segmentation head decodes the VLM’s output token into a mask. A key challenge is that material selection datasets with text annotations are unavailable. We therefore propose a scalable data generation pipeline: we collect real and synthetic images with material masks, then leverage VLMs to generate material descriptions with rich visual-semantic information. Using the generated data, we train MAOAM with a multi-task objective over click- and text-based selection, along with an auxiliary VQA task derived from the material descriptions to facilitate deeper material understanding. Despite being trained with uni-modal prompts, our model exhibits an emergent improvement in selection quality when combining text and clicks at inference time, enabling more flexible image editing workflows. Experiments demonstrate accurate and coherent selections across diverse objects, materials, and interaction scenarios, highlighting robustness in practice.
Representing complex geometry as a displacement function defined over a coarse base mesh enables compact storage and accelerated rendering. The core challenge in converting detailed triangle meshes into this representation is computing base meshes that have as few triangles as possible, while also supporting displacement functions that accurately approximate the input. Accurate approximation requires the supported displacement functions to bijectively map the input surface onto the base with low parametric distortion. We observe that this distortion can be measured by evaluating the pointwise Jacobian of the displacement functions. Our new DJM (Displacement Jacobian Metric)-based base-mesh construction method uses the Jacobian of the displacement functions to guide base mesh computation, enabling us to outperform prior approaches in terms of accuracy to size trade-off. We achieve this goal by proposing a variant of the QEM-based simplification scheme that constrains the displacement mapping between the input and the base to be bijective and low distortion (defined as satisfying a lower bound on the mapping Jacobian). When evaluating and encoding the displacement maps, we avoid unreliable ray-mesh intersections by explicitly storing the mapping between the input mesh and the base throughout the construction process, and use this mapping within a robust inverse barycentric displacement solver to obtain dense base-to-mesh correspondences to assist all computations. We demonstrate DJM to outperform alternative schemes in terms of reconstruction accuracy to size trade-off, and demonstrate its robustness and usability for micromesh based rendering and neural encoding.
Recent radiance-field-based reconstruction methods, such as NeRF and 3DGS, achieve high visual fidelity for indoor scenes, but often break down under scene editing due to baked illumination and the lack of explicit light transport. In contrast, inverse path tracing methods based on mesh representations enforce correct light transport but require highly accurate geometry, making them difficult to apply robustly to real indoor scenes. We present Emission-Aware Gaussians and Path Tracing (EAG-PT), a method for physically based reconstruction and rendering of indoor scenes using a unified 2D Gaussian representation, targeting editable diffuse global illumination. Our approach consists of three key ideas: (1) representing indoor scenes with 2D Gaussians as a transport-friendly geometric proxy that avoids explicit mesh reconstruction; (2) explicitly separating emissive and non-emissive components during reconstruction to support editing; and (3) decoupling reconstruction from final rendering by using efficient single-bounce optimization and high-quality multi-bounce path tracing, respectively. Experiments on synthetic and real indoor scenes show that EAG-PT produces more natural and physically consistent edited renderings than radiance-field reconstructions, while preserving finer geometric detail and avoiding mesh-induced artifacts compared with mesh-based inverse path tracing. These results highlight the potential of our approach for applications such as interior design, XR content creation, and embodied AI.
Computing injective, flip-free embeddings of simplicial meshes is a fundamental challenge in geometry processing. While the Total Lifted Content (TLC) energy provides a smooth functional with theoretically guaranteed injective global minima, its non-convex nature presents significant optimization hurdles. In this work, we propose a Nonlinear Multigrid (MG/Opt) framework specifically tailored for efficient TLC energy minimization. Our approach leverages a topology-preserving mesh hierarchy to rapidly eliminate low-frequency errors on coarse scales while maintaining rigorous gradient consistency between hierarchical levels through linear correction terms. To address the persistent challenge of local inversions, we introduce a targeted smoothing strategy and an adaptive parameter scheduling mechanism that balances global convergence speed with local injectivity precision. Experimental evaluations across extensive 2D and 3D benchmark datasets demonstrate the robustness of our approach. Our method achieves a 100% convergence success rate across all tested 2D cases and all but one 3D dataset (the VOLMAP benchmark). Furthermore, our MG/Opt framework delivers substantial performance gains, achieving up to 63 × speedup on high-resolution subdivided datasets compared to traditional solvers. These results confirm that our multigrid strategy provides a scalable and reliable solution for large-scale injective mapping problems. Code and data for this paper are at https://github.com/distmeshM/MGOpt_TLC
Geometry optimization often encounters sparse gradients from limited observations, which can cause optimization to drift toward unnatural shapes. Prior work stabilizes this process by exploiting spatial structure encoded by the Laplacian operator within a single shape, enforcing spatial smoothness on gradients. However, such smoothness alone propagates gradients to unobserved regions without any prior knowledge of how shapes typically deform in a given domain. We introduce statistical gradient filtering, which leverages statistical structure across a shape collection by learning shape variations via principal component analysis (PCA) and guiding geometry updates along directions consistent with this learned prior. Unlike previous PCA-based methods that constrain solutions to a linear subspace or modify the objective with regularization terms, we filter gradients at each iteration to steer the optimization path toward plausible shapes, without restricting the solution space or altering the original objective. We validate our approach across a range of shape optimization tasks, demonstrating robust convergence even under challenging conditions.
We present a training algorithm to mitigate optimization instabilities in small neural networks, like those used in real-time neural shading applications. While large, overparameterized models exhibit predictable convergence, smaller architectures often suffer from high optimization variance: differently initialized models converge to disparate local minima. To reduce these training instabilities, we introduce an optimization approach that utilizes an ensemble of network instances during training. We prune underperforming instances and dynamically resize training batches to maintain wall-clock timings comparable to—or faster than—single-instance training. This strategy allows efficiently exploring the weight space yielding a well-performing model with significantly higher likelihood than optimizing a single instance only. We develop and analyze the algorithm in the context of learning reflectance functions for neural shading. This is a challenging task for small neural models due to the high dynamic range of the target function. In addition to our multi-instance training method, we also revisit the choices of loss functions, activation functions, and input parameterization to further improve quality and training robustness.
Recovering high-quality video from sparse event streams is a challenging task. Regression methods often blur textures, while existing generative models struggle with long-term stability. We propose LongE2V, a novel approach that leverages pre-trained video diffusion priors to jointly handle event-based video reconstruction, prediction, and frame interpolation. By fine-tuning a foundational video model, our approach achieves high data efficiency and superior perceptual quality. We introduce Autoregressive Unrolling and Adaptive Context Switching to mitigate temporal drift in extremely long sequences. We also propose Reencoding Alignment with Cross Residual Correction to ensure precise bidirectional consistency during frame interpolation. Furthermore, Event Voxel Density Augmentation ensures robustness across varying sensor resolutions. Extensive experiments on real-world benchmarks demonstrate that LongE2V outperforms state-of-the-art methods across all three tasks, exhibiting exceptional temporal coherence and zero-shot generalization.
Existing volumetric capture of dynamic human performance achieves high fidelity with dense camera arrays. However, in real-world scenarios, only a handful of low-overlap cameras are available, which degrades the output quality and leaves large areas unobserved. Recent 4D reconstruction methods have focused on low-overlap settings, yet they still produce noticeable artifacts in under-observed regions. Video diffusion models have emerged as another option, but they show geometrically inconsistent results for humans. To address these limitations, we propose StudioRecon, a pipeline that reconstructs 4D human scenes from sparse, low-overlap cameras by decoupling background and humans. We densify background supervision by synthesizing hundreds of camera-controlled novel views with a video diffusion model. We also robustly initialize deformable Gaussian humans with cross-view identity association and triangulated multi-view keypoint fitting. Finally, our recursive enhancement module with motion-adaptive consistency injection harmonizes the composed output, thereby further avoiding remaining artifacts. We achieve state-of-the-art novel view synthesis across four real-world datasets and demonstrate applications such as novel trajectory rendering and human replacement. Project page: https://sisyphm.github.io/studiorecon-page/.