English(EN)EVA01: Unified Native 3D Understanding and Generation via Mixture-of-Transformers
AI研究推动3D重建和场景理解进展
作者PulseAugur 编辑部·[76 个来源]·
研究人员正在探索用于3D重建和场景理解的先进技术，重点关注优化计算资源和提高准确性。研究探讨了医学成像中2D、2.5D和3D模型之间的权衡，研究结果表明2.5D CNN提供了有利的平衡。其他工作引入了用于扩散时间步长调度的框架，以提高3D CT重建的效率和保真度。此外，正在开发新的在线3D视觉-语言模型，用于从流式视频进行实时空间理解，并提出了自适应特征优化方法来提高3D场景重建的质量。
AI
Most recent 3D reconstruction and editing systems operate on implicit and explicit representations such as NeRF, point clouds, or meshes. While these representations enable high-fidelity rendering, they are fundamentally low-level and hard to control programmatically. In contrast…
arXiv cs.AI
TIER_1English(EN)·Md Enamul Hoq, Sharafat Hossain, Imraul Emmaka, Linda Larson-Prior, Lawrence Tarbox, Jonathan Bona, Donald Johann Jr. and Fred Prior·
arXiv:2606.06950v1 Announce Type: cross Abstract: Three-dimensional models are widely assumed preferable for volumetric medical imaging, yet their practical value depends on whether performance gains justify added computational cost and complexity. Rather than proposing a new arc…
arXiv:2606.06236v1 Announce Type: new Abstract: Pretrained diffusion models demonstrate impressive potential in solving highly ill-posed 3D computed tomography (CT) inverse problems, while the inference process suffers from significant computational overhead. Furthermore, existin…
An online 3D vision-language model enables real-time spatial understanding from streaming video using autoregressive control modeling and efficient visual token compression.
Pretrained diffusion models demonstrate impressive potential in solving highly ill-posed 3D computed tomography (CT) inverse problems, while the inference process suffers from significant computational overhead. Furthermore, existing uniform timestep schedules fail to capture the…
arXiv cs.AI
TIER_1English(EN)·Samuel Garcin, Thomas Walker, Steven McDonagh, Tim Pearce, Hakan Bilen, Tianyu He, Kaixin Wang, Jiang Bian·
arXiv:2603.03482v2 Announce Type: replace-cross Abstract: Interactive world models continually generate video by responding to a user's actions, enabling open-ended generation capabilities. However, existing models typically lack a 3D representation of the environment, meaning 3D…
arXiv:2604.09877v2 Announce Type: replace-cross Abstract: At the intersection of computer vision and robotic perception, 4D reconstruction of dynamic scenes connects low-level geometric sensing with high-level semantic understanding. We present Genie 4D, a framework that turns ha…
arXiv cs.AI
TIER_1English(EN)·Xihang Yu, Rajat Talak, Lorenzo Shaikewitz, Luca Carlone·
arXiv:2602.08058v3 Announce Type: replace-cross Abstract: In the presence of occlusions and measurement noise, geometrically accurate scene reconstructions -- which fit the sensor data -- can still be physically incorrect. For instance, when estimating the poses and shapes of obj…
arXiv:2505.08438v4 Announce Type: replace-cross Abstract: Event cameras are rapidly emerging as powerful vision sensors for 3D reconstruction, uniquely capable of asynchronously capturing per-pixel brightness changes. Compared to traditional frame-based cameras, event cameras pro…
3D vision research is organized through a taxonomy connecting geometric representations, datasets, learning frameworks, and applications across reconstruction, generation, and video modeling tasks.
arXiv:2605.31534v1 Announce Type: cross Abstract: Three-dimensional scene reconstruction depends on local image evidence that is both visually discriminative and geometrically useful. Fixed feature thresholds and uniform feature budgets are easy to deploy, but they can waste comp…
Three-dimensional scene reconstruction depends on local image evidence that is both visually discriminative and geometrically useful. Fixed feature thresholds and uniform feature budgets are easy to deploy, but they can waste computation on repeated texture, low-parallax regions,…
arXiv:2605.30310v1 Announce Type: cross Abstract: City-scale 3D surface reconstruction from multiview images for downstream 3D simulation, poses highly challenging problems due to the scale and complexity of urban scenes. Existing city-scale 3D reconstruction methods based on NeR…
City-scale 3D surface reconstruction from multiview images for downstream 3D simulation, poses highly challenging problems due to the scale and complexity of urban scenes. Existing city-scale 3D reconstruction methods based on NeRF, Gaussian Splatting etc. often fail to recover 3…
ViGeo is a transformer-based foundation model that recovers dense and consistent 3D geometry from videos using dynamic chunking attention and a completion-based data refinement framework.
Category-level 3D correspondence is learned from single images through a shared morphable object prior, enabling semantic 3D object understanding without explicit correspondence supervision.
TriSplat is a feed-forward 3D reconstruction network that uses oriented triangle primitives to directly generate simulation-ready meshes from single images, bypassing expensive post-processing steps.
A novel diffusion-based framework for multi-view 3D reconstruction that restores both scene geometry and high-quality imagery from degraded inputs by operating in the feature space of a 3D reconstructor.
A novel method for 3D scene reconstruction that integrates generative 3D priors with multi-view image conditioning to produce high-fidelity, editable mesh reconstructions of indoor environments.
EVA01 enables native 3D mesh integration in multimodal language models through a Mixture-of-Transformers architecture that aligns semantic and geometric manifolds for improved generation and editing capabilities.
arXiv:2606.10142v1 Announce Type: new Abstract: Recent advances in 3D generation have led to substantial improvements in realism, controllability, and efficiency, yet the evaluation of 3D assets remains underexplored. Existing evaluation paradigms, including human evaluation, lea…
arXiv:2606.10478v1 Announce Type: new Abstract: Most recent 3D reconstruction and editing systems operate on implicit and explicit representations such as NeRF, point clouds, or meshes. While these representations enable high-fidelity rendering, they are fundamentally low-level a…
arXiv:2606.11152v1 Announce Type: new Abstract: Multimodal large language models can write code to produce complex programs as well as use programs to do 3D modeling, which opens up a new avenue for 3D generation powered by their priors, world knowledge and reasoning. Yet existin…
arXiv:2504.18424v2 Announce Type: replace Abstract: We present Layered Ray Intersections (LaRI), a fully supervised method for occluded geometry reasoning from a single image. Unlike conventional depth estimation, which is limited to visible surfaces, LaRI predicts multiple surfa…
Multimodal large language models can write code to produce complex programs as well as use programs to do 3D modeling, which opens up a new avenue for 3D generation powered by their priors, world knowledge and reasoning. Yet existing benchmarks rarely evaluate 3D modeling through…
Most recent 3D reconstruction and editing systems operate on implicit and explicit representations such as NeRF, point clouds, or meshes. While these representations enable high-fidelity rendering, they are fundamentally low-level and hard to control programmatically. In contrast…
arXiv:2606.08980v1 Announce Type: new Abstract: This paper introduces EPS3D, a new end-to-end feed-forward framework for open-vocabulary 3D panoptic segmentation. Unlike existing methods relying on additional preprocessing, we design an end-to-end architecture, with a distillatio…
arXiv cs.CV
TIER_1English(EN)·Hanxun Yu, Xuan Qu, Lei Ke, Boqiang Zhang, Yuxin Wang, Jianke Zhu, Dong Yu·
arXiv:2606.06891v1 Announce Type: new Abstract: Despite advances in 3D scene understanding, existing 3D Large Multimodal Models operate in offline settings, requiring complete scene observations or predefined video clips. In this paper, we present an online 3D vision-language mod…
arXiv cs.CV
TIER_1English(EN)·Donald Johann Jr. and Fred Prior·
Three-dimensional models are widely assumed preferable for volumetric medical imaging, yet their practical value depends on whether performance gains justify added computational cost and complexity. Rather than proposing a new architecture, we study how input dimensionality (2D, …
Despite advances in 3D scene understanding, existing 3D Large Multimodal Models operate in offline settings, requiring complete scene observations or predefined video clips. In this paper, we present an online 3D vision-language model that enables real-time spatial understanding …
arXiv:2606.06485v1 Announce Type: new Abstract: Recent advances in 3D multimodal large language models (3D-MLLMs) have enabled unified solutions for 3D scene understanding tasks, including visual question answering, captioning, and referring segmentation. However, existing 3D-MLL…
arXiv:2507.12336v2 Announce Type: replace Abstract: Most existing 3D keypoint estimation methods rely on manual annotations or calibrated multi-view images, both of which are expensive to collect. This paper introduces KeyDiff3D, a framework that can accurately predict 3D keypoin…
Recent advances in 3D multimodal large language models (3D-MLLMs) have enabled unified solutions for 3D scene understanding tasks, including visual question answering, captioning, and referring segmentation. However, existing 3D-MLLMs remain largely object-centric, limiting their…
arXiv:2606.05035v1 Announce Type: new Abstract: Long-horizon online visual mapping is a core capability for robot perception, requiring continuous camera-motion and scene-geometry estimation from visual streams under bounded memory and computation. Recent feed-forward 3D reconstr…
arXiv:2606.04291v1 Announce Type: new Abstract: 3D vision has rapidly evolved, driven by increasingly diverse data representations, learning paradigms, and modeling strategies. Yet the field remains fragmented across representations and benchmarks, making it difficult to develop …
arXiv:2606.04593v1 Announce Type: new Abstract: Although dynamic 3D (i.e., 4D) reconstruction from a monocular dynamic camera has recently advanced, it remains fundamentally limited by depth ambiguity. In this paper, we focus on an alternative practical way, i.e., sparse dynamic …
arXiv:2606.04871v1 Announce Type: new Abstract: The selection of an appropriate 3D representation is a fundamental design decision that dictates the efficiency, quality, and capabilities of modern computer vision and graphics pipelines for tasks such as 3D reconstruction, novel-v…
arXiv:2606.04891v1 Announce Type: new Abstract: Generating compact polygonal models from point clouds is a key problem in 3D vision and computer graphics. However, due to inherent limitations of LiDAR scanning (e.g. range constraints and occlusions), critical scene information is…
Long-horizon online visual mapping is a core capability for robot perception, requiring continuous camera-motion and scene-geometry estimation from visual streams under bounded memory and computation. Recent feed-forward 3D reconstruction models provide strong geometric priors, b…
Generating compact polygonal models from point clouds is a key problem in 3D vision and computer graphics. However, due to inherent limitations of LiDAR scanning (e.g. range constraints and occlusions), critical scene information is often missing, leading to degraded reconstructi…
The selection of an appropriate 3D representation is a fundamental design decision that dictates the efficiency, quality, and capabilities of modern computer vision and graphics pipelines for tasks such as 3D reconstruction, novel-view synthesis and rendering, shape and motion an…
Although dynamic 3D (i.e., 4D) reconstruction from a monocular dynamic camera has recently advanced, it remains fundamentally limited by depth ambiguity. In this paper, we focus on an alternative practical way, i.e., sparse dynamic camera setup, where a handful of independently m…
arXiv:2406.18544v4 Announce Type: replace Abstract: 3D Gaussian Splatting (3DGS) has shown a powerful capability for novel view synthesis due to its detailed expressive ability and highly efficient rendering speed. Unfortunately, creating relightable 3D assets and reconstructing …
arXiv cs.CV
TIER_1English(EN)·Ranran Huang, Weixun Luo, Ye Mao, Krystian Mikolajczyk·
arXiv:2603.27455v2 Announce Type: replace Abstract: In this paper, we introduce NAS3R, a self-supervised feed-forward framework that jointly learns explicit 3D geometry and camera parameters with no ground-truth annotations and no pretrained priors. During training, NAS3R reconst…
arXiv:2603.14377v2 Announce Type: replace Abstract: Prevailing High Dynamic Range (HDR) video reconstruction methods are fundamentally trapped in a fragile alignment-and-fusion paradigm. While explicit spatial alignment can successfully recover fine details in controlled environm…
arXiv cs.CV
TIER_1English(EN)·Inhee Lee, Sangwon Baik, Sungjoo Kim, Hyeonwoo Kim, Hyunsoo Cha, Hanbyul Joo·
arXiv:2606.03994v1 Announce Type: new Abstract: Reconstructing interactive, simulation-ready 3D scenes from a single image is a critical bottleneck for robotic manipulation. While recent single-image lifters recover plausible per-object shapes, composing them yields scenes that c…
Reconstructing interactive, simulation-ready 3D scenes from a single image is a critical bottleneck for robotic manipulation. While recent single-image lifters recover plausible per-object shapes, composing them yields scenes that collapse under physical simulation due to interpe…
arXiv:2606.01367v1 Announce Type: cross Abstract: Active scene reconstruction enables robots/UAVs to autonomously plan trajectories and reconstruct environments without costly manual data acquisition. Unlike passive methods, active reconstruction requires real-time construction o…
arXiv cs.CV
TIER_1English(EN)·Sebastian Koch, Johanna Wald, Hidenobu Matsuki, Pedro Hermosilla, Timo Ropinski, Federico Tombari·
arXiv:2512.14364v3 Announce Type: replace Abstract: Holistic 3D scene understanding involves capturing and parsing unstructured 3D environments. Due to the inherent complexity of the real world, existing models have predominantly been developed and limited to be task-specific. We…
arXiv cs.CV
TIER_1English(EN)·Adrian Ramlal, John S. Zelek·
arXiv:2606.00452v1 Announce Type: new Abstract: Dynamic scene reconstruction via 3D Gaussian Splatting (3DGS) has emerged as a compelling approach for representing evolving environments, yet understanding trade-offs between methodologies remains crucial. This paper presents a com…
arXiv:2507.23277v3 Announce Type: replace Abstract: Feed-forward 3D modeling has emerged as a promising approach for rapid and high-quality 3D reconstruction. In particular, directly generating explicit 3D representations, such as 3D Gaussian splatting, has attracted significant …
arXiv:2605.31466v1 Announce Type: new Abstract: Reconstructing the complete geometry of a scene from a single RGB image remains challenging - especially when inferring hidden structures where visual evidence is incomplete. We introduce VolFill, a generative framework that predict…
arXiv:2605.21472v2 Announce Type: replace Abstract: View-conditioned 3D generators such as SAM 3D, TRELLIS, and Hunyuan3D produce high-quality object reconstructions from a single view, but real-world visual observation often arrives as long monocular streams. Naively applying th…
Reconstructing the complete geometry of a scene from a single RGB image remains challenging - especially when inferring hidden structures where visual evidence is incomplete. We introduce VolFill, a generative framework that predicts the 3D structure of the complete scene rather …
arXiv cs.CV
TIER_1English(EN)·Alessandro Burzio, Tobias Fischer, Sven Elflein, Qunjie Zhou, Riccardo de Lutio, Jiawei Ren, Jiahui Huang, Shengyu Huang, Marc Pollefeys, Laura Leal-Taix\'e, Zan Gojcic, Haithem Turki·
arXiv:2605.30215v1 Announce Type: new Abstract: Recent feed-forward 3D reconstruction transformers have scaled to over a billion parameters, following the broader trend of increasing model capacity in computer vision. Yet emerging evidence suggests that contiguous transformer lay…
arXiv cs.CV
TIER_1English(EN)·Daniel Rho, Jun Myeong Choi, Matthew Thornton, Biswadip Dey, Roni Sengupta·
arXiv:2605.30320v1 Announce Type: new Abstract: Existing inverse physics methods recover physical parameters from multi-view videos, where geometric constraints across views resolve scale and 3D structure. In monocular settings, however, such constraints are absent, leading to se…
arXiv:2605.30338v1 Announce Type: new Abstract: Reconstructing physically stable 3D scenes from a single RGB image enables casual images to be converted into simulation-ready digital assets for applications such as immersive interaction and content creation. However, existing sin…
arXiv:2605.30060v1 Announce Type: new Abstract: This work presents ViGeo, a feed-forward foundation model for recovering spatially dense and temporally consistent geometry from video sequences. Built upon a plain transformer architecture without task-specific architectural modifi…
Reconstructing physically stable 3D scenes from a single RGB image enables casual images to be converted into simulation-ready digital assets for applications such as immersive interaction and content creation. However, existing single-image reconstruction methods fall short in c…
Existing inverse physics methods recover physical parameters from multi-view videos, where geometric constraints across views resolve scale and 3D structure. In monocular settings, however, such constraints are absent, leading to severe scale ambiguity, inaccurate geometry, and w…
Recent feed-forward 3D reconstruction transformers have scaled to over a billion parameters, following the broader trend of increasing model capacity in computer vision. Yet emerging evidence suggests that contiguous transformer layers often behave like repeated applications of s…
This work presents ViGeo, a feed-forward foundation model for recovering spatially dense and temporally consistent geometry from video sequences. Built upon a plain transformer architecture without task-specific architectural modifications, ViGeo supports streaming, full-sequence…
arXiv:2605.28125v1 Announce Type: new Abstract: Many real-world 3D reconstruction applications demand photorealism and metric accuracy across unbounded, complex scenes with challenging lighting and imperfect captures that current Neural Radiance Field (NeRF) pipelines only partly…
arXiv cs.CV
TIER_1English(EN)·Leonhard Sommer, Artur Jesslen, Basavaraj Sunagad, Adam Kortylewski·
arXiv:2605.28257v1 Announce Type: new Abstract: Understanding 3D objects from images is fundamental to robotics and AR/VR applications. While recent work has made progress in category-level pose estimation, current representations fail to capture the fine-grained semantics needed…
arXiv cs.CV
TIER_1English(EN)·Haitang Feng, Xinkai Chen, Jie Liu, Jie Tang, Gangshan Wu, Beiqi Chen, Jianhuang Lai, Guangcong Wang·
arXiv:2508.18271v2 Announce Type: replace Abstract: 3D object inpainting is commonly achieved via multi-view 2D image completion, yet independently inpainted views often suffer from cross-view inconsistencies, leading to blurred textures, geometric discontinuities, and visual art…
Understanding 3D objects from images is fundamental to robotics and AR/VR applications. While recent work has made progress in category-level pose estimation, current representations fail to capture the fine-grained semantics needed for reasoning about object parts, functions, an…
Many real-world 3D reconstruction applications demand photorealism and metric accuracy across unbounded, complex scenes with challenging lighting and imperfect captures that current Neural Radiance Field (NeRF) pipelines only partly satisfy. This study adapts NeRF-based 3D recons…
arXiv cs.CV
TIER_1English(EN)·Jin Hyeon Kim, Jaeeun Lee, Claire Kim, Kyoungjin Oh, Paul Hyunbin Cho, Jaewon Min, Yeji Choi, Jihye Park, Hyunhee Park, Minkyu Park, Seungryong Kim·
arXiv:2605.26230v1 Announce Type: new Abstract: Multi-view 3D reconstruction has achieved remarkable progress with the advent of feed-forward 3D reconstruction models. However, these models are typically trained and evaluated under ideal, degradation-free imaging conditions, wher…
arXiv:2605.26519v1 Announce Type: new Abstract: Recent feed-forward geometry foundation models have demonstrated impressive generalization by recovering depth and poses in a single forward pass. However, these models are typically constrained by a global coordinate frame assumpti…
arXiv cs.CV
TIER_1English(EN)·Weijie Wang, Zimu Li, Jinchuan Shi, Zeyu Zhang, Botao Ye, Marc Pollefeys, Donny Y. Chen, Bohan Zhuang·
arXiv:2605.26115v1 Announce Type: new Abstract: Sparse-view 3D reconstruction is increasingly addressed with feed-forward splatting networks that predict explicit primitives directly from images. Yet most existing methods remain centered on Gaussian primitives and expose surfaces…
arXiv cs.CV
TIER_1English(EN)·Wanhee Lee, Klemen Kotar, Rahul Mysore Venkatesh, Jared Watrous, Honglin Chen, Khai Loong Aw, Daniel L. K. Yamins·
arXiv:2605.24321v1 Announce Type: new Abstract: Understanding 3D scenes requires flexible combinations of visual reasoning tasks, including depth estimation, novel view synthesis, and object manipulation, all of which are essential for perception and interaction. Existing approac…
Sparse-view 3D reconstruction is increasingly addressed with feed-forward splatting networks that predict explicit primitives directly from images. Yet most existing methods remain centered on Gaussian primitives and expose surfaces only indirectly: extracting a usable mesh for d…
arXiv cs.CV
TIER_1English(EN)·Katharina Schmid, Nicolas von L\"utzow, Jozef Hladk\'y, Angela Dai, Matthias Nie{\ss}ner·
arXiv:2605.23888v1 Announce Type: new Abstract: We introduce a new approach to high-fidelity 3D scene reconstruction from multi-view RGB images that tightly couples reconstruction with a strong generative 3D prior. We cast scene reconstruction as conditional 3D generation over a …
arXiv:2605.22997v1 Announce Type: new Abstract: In autonomous driving, mapping is critical for motion planning but remains an under-utilized resource for perception tasks such as 3D object detection. Maps can provide robust structural priors of the static environment, helping res…
We introduce a new approach to high-fidelity 3D scene reconstruction from multi-view RGB images that tightly couples reconstruction with a strong generative 3D prior. We cast scene reconstruction as conditional 3D generation over a set of spatially-localized, overlapping chunks t…
<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WrRe6mLNWqdOKKNQxuE6bQ.png" /></figure><h4><strong><em>A hands-on benchmark, a COLMAP comparison, and a full TensorRT FP16 conversion of a 1.26B-parameter 3D reconstruction Transformer.</em></strong></h4><p>If yo…
