Computer Vision

Home /
Categories /
Computer Vision

HuSc3D: Human Sculpture Dataset for 3D Object Reconstruction

Abstract 3D scene reconstruction from 2D images is important tasks in computer graphics. Unfortunately, existing datasets and benchmarks concentrate on idealized synthetic or meticulously captured realistic data. Such benchmarks fail to convey the inherent complexities encountered in newly acquired real-world scenes. In such scenes the background is often dynamic, and by popular usage of cell phone cameras, there might be discrepancies in, e.g., white balance. To address this gap, we present HuSc3D, a novel dataset specifically designed for rigorous benchmarking of 3D reconstruction models under realistic acquisition challenges. Our dataset features six highly detailed, fully white sculptures characterized by intricate perforations and minimal textural variation. Furthermore, the number of images per scene varies significantly, introducing the additional challenge of limited training data for some instances alongside scenes with a standard number of views. By evaluating 3D reconstruction methods on this diverse dataset, we demonstrate the distinctiveness of HuSc3D in effectively differentiating model performance, particularly highlighting the sensitivity of methods to fine geometric details, color ambiguity, and varying data availability – limitations often masked by more conventional datasets.

MeshSplats: Mesh-Based Rendering with Gaussian Splatting Initialization

Abstract Gaussian Splatting (GS) is an emerging, pivotal technique in 3D computer graphics. Most GS-based algorithms exclude the use of classical methods such as ray tracing, which offer numerous inherent advantages in rendering. For example, ray tracing can handle incoherent rays for advanced lighting effects, including shadows and reflections. To address this issue, we introduce MeshSplats, a method which converts GS to a mesh-like format. Following the completion of training, MeshSplats transforms Gaussian elements into mesh faces, enabling rendering using ray tracing methods with all their associated benefits. Our model can be used immediately after transformation, yielding a mesh without additional training, with only a slight reduction in reconstruction quality. We can enhance the quality by applying a dedicated optimization algorithm that operates on mesh faces rather than Gaussian components. Importantly, MeshSplats acts as a wrapper, converting pre-trained GS models into a ray-traceable format. The efficacy of our method is substantiated by experimental results, underscoring its extensive applications in computer graphics and image processing.

LumiMotion: Improving Gaussian Relighting with Scene Dynamics

Abstract In 3D reconstruction, the problem of inverse rendering, namely recovering the illumination of the scene and the material properties, is fundamental. Existing Gaussian Splatting-based methods primarily target static scenes and often assume simplified or moderate lighting to avoid entan- gling shadows with surface appearance. This limits their ability to accurately separate lighting effects from mate- rial properties, particularly in real-world conditions. We address this limitation by leveraging dynamic elements– regions of the scene that undergo motion–as a supervisory signal for inverse rendering. Motion reveals the same sur- faces under varying lighting conditions, providing stronger cues for disentangling material and illumination. This the- sis is supported by our experimental results which show we improve LPIPS by 23% for albedo estimation and by 15% for scene relighting relative to next-best baseline. To this end, we introduce LumiMotion, the first Gaussian-based approach that leverages dynamics for inverse rendering and operates in arbitrary dynamic scenes. Our method learns a dynamic 2D Gaussian Splatting representation that em- ploys a set of novel constraints which encourage the dy- namic regions of the scene to deform, while keeping static regions stable. As we demonstrate, this separation is crucial for correct optimization of the albedo. Finally, we release a new synthetic benchmark comprising five scenes under four lighting conditions, each in both static and dynamic variants, for the first time enabling systematic evaluation of inverse rendering methods in dynamic environments and challenging lighting.

EPIC: Explanation of Pretrained Image Classification Networks via Prototypes

Abstract Explainable AI (XAI) methods generally fall into two categories. Post-hoc approaches generate explanations for pre-trained models and are compatible with various neural network architectures. These methods often use feature importance visualizations, such as saliency maps, to indicate which input regions influenced the model’s prediction. Unfortunately, they typically offer a coarse understanding of the model’s decision-making process. In contrast, ante-hoc (inherently explainable) methods rely on specially designed model architectures trained from scratch. A notable subclass of these methods provides explanations through prototypes, representative patches extracted from the training data. However, prototype-based approaches require dedicated architectures, involve specialized training procedures, and perform well only on specific datasets. In this work, we propose EPIC (Explanation of Pretrained Image Classification), a novel approach that bridges the gap between these two paradigms. Like post-hoc methods, EPIC operates on pre-trained models without architectural modifications. Simultaneously, it delivers intuitive, prototype-based explanations inspired by ante-hoc techniques. To the best of our knowledge, EPIC is the first post-hoc method capable of fully replicating the core explanatory power of inherently interpretable models. We evaluate EPIC on benchmark datasets commonly used in prototype-based explanations, such as CUB-200-2011 and Stanford Cars, alongside large-scale datasets like ImageNet, typically employed by post-hoc methods. EPIC uses prototypes to explain model decisions, providing a flexible and easy-to-understand tool for creating clear, high-quality explanations.

CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting

Abstract We present CLIPGaussian, a framework that performs universal, multimodal style transfer by operating directly on Gaussian primitives—the shared representation underlying modern 2D, 3D, and 4D Gaussian Splatting pipelines. Instead of relying on heavy generative models, CLIPGaussian uses CLIP-driven text and image cues to steer both the appearance and, when beneficial, the geometry of Gaussian elements. This makes stylization a geometric-aware process: colors shift in locally meaningful ways, textures reorganize to match the target style, and structural details adapt subtly without disrupting the underlying scene. Because the method works natively on Gaussian fields, the same mechanism seamlessly applies to single images, complex 3D objects, temporally coherent videos, and dynamic 4D reconstructions.

MiraGe: Editable 2.5D Image Representations with Flat-Controlled 3D Gaussians

Abstract Implicit Neural Representations (INRs) encode images as continuous functions that map pixel coordinates to RGB values, achieving compact storage and high visual fidelity. Recent work such as GaussianImage replaces neural MLPs with collections of 2D Gaussian primitives, reaching similar reconstruction quality and compression but offering limited editability. In practice, creators often need to adjust content—move objects, bend a photo, cast new shadows, or create parallax—all of which are awkward within purely 2D or purely additive Gaussian schemes.

VeGaS: Video Gaussian Splatting

Abstract Modern video representations often prioritize reconstruction quality and compression, but they can be difficult to edit in a precise and controllable way. VeGaS (Video Gaussian Splatting) tackles this gap by adapting Gaussian Splatting ideas to 2D videos while explicitly modeling nonlinear motion and appearance changes across time.