The paper introduces SightCAD, a framework for parametric CAD reconstruction that explicitly handles view feasibility constraints in realistic indoor scenes. It learns a view selector to choose a small set of feasible viewpoints for a vision‑language model that generates executable CadQuery code, and jointly trains the selector and the VLM using a reward that depends on the final geometric fidelity. The approach outperforms baseline methods on several benchmarks, achieving higher intersection‑over‑union scores and a lower invalid‑code rate.
By Kartik Bali, Mahish Guru, Yiderigun Borjigin, Alexandra Starostina, Christian J. Cyron, Roland Aydin
CADSplat is a framework that reconstructs photorealistic, geometrically accurate digital twins from fewer than 15 wide‑baseline images by regularizing 3D Gaussian Splatting with an explicit CAD shape prior. It matches segmented object silhouettes to a CAD library to retrieve a suitable model and camera poses, then anchors Gaussian primitives to the model’s surface and jointly optimizes splat parameters, registration, and a non‑rigid deformation field. Experiments on two real‑world datasets show CADSplat outperforms baselines, especially in sparse and self‑occluded scenarios, and its gains mainly stem from constraining splats to a surface rather than the CAD shape itself.
By Kristof Overdulve, Lode Jorissen, Nick Michiels
CALIPER is a model‑free RGB‑D framework that performs fine‑grained recognition of visually similar industrial parts by combining support‑based appearance matching with metric size evidence. Each class is onboarded from a single turntable RGB‑D video and a few labeled real images, enabling 3D reconstruction for appearance support and depth‑aligned size profiling. At inference, a YOLOv8n‑seg model localizes parts, a frozen DINOv2 backbone with an episodically trained embedding head matches support, and margin‑conditioned metric fusion selectively uses size evidence for ambiguous cases, achieving high accuracy on 18 parts and robust enrollment of unseen screws without retraining.
By Alankrit Gupta, Chenxi Tao, Seung-Kyum Choi
AnyBox is a zero‑shot framework that estimates the full 9DoF pose (6D pose plus 3D dimensions) of boxes from a single RGB‑D image, leveraging the geometric regularity of boxes. It alternates between pose and scale estimation, using a binary search guided by the discrepancy between a reprojected template and the observed mask, and employs a depth‑consistency filter and an early‑stopping rule to prune implausible hypotheses. On public benchmarks and a warehouse dataset, AnyBox improves detection AP by up to 36 points and boosts robotic box‑shelving success by 28%.
By Yintao Ma, Sajjad Pakdamansavoji, Charles Eret, Rui Heng Yang, Xuan Zhao, Yingxue Zhang, Tongtong Cao, Amir Rasouli
The study investigates how scan-derived class prototypes influence recognition performance when used as a fixed output layer in a recognizer. By systematically varying prototype content (real, random, permuted), supervision, and query protocols across datasets such as T‑LESS, HOPE, and industrial parts, the authors find that real prototypes provide only a modest accuracy gain—up to one percentage point—over controls, with larger differences sometimes favoring the controls. The analysis also shows that matched logistic regression can approach or surpass the fused prototype performance, and that background protocols significantly affect prototype accuracy.
By Chenxi Tao, Hong-In Won, Seung-Kyum Choi
arXiv:2605. 30581v2 Announce Type: replace-cross Abstract: Industrial visual sim-to-real is often described as transferring from synthetic images to real images, but industrial deployment usually involves a broader mismatch between available evidence and required decisions.
By Chenxi Tao, Seung-Kyum Choi
Matisse is a training‑free framework that combines active 3D reconstruction with keyframe selection by using evidence from a pretrained generative 3D model. It estimates evidential uncertainty via cross‑attention on 3D latent tokens and derives an evidential information gain to guide view acquisition and keyframe selection, reducing redundant observations and supporting multi‑object scenes with occlusion‑aware aggregation. On GSO30, YCB‑V, and Replica, Matisse improves Chamfer distance by 12.7%, 3.8%, and 9.2% respectively, and speeds up end‑to‑end reconstruction by 1.5× compared to the best baseline.
By Xihang Yu, Kaichen Zhou, Lorenzo Shaikewitz, Cl\'ement Jambon, Xiao Zhan, Rajat Talak, Luca Carlone
OmniAct3D is a framework that adapts Vision Foundation Model (VFM) detectors trained on perspective images to equirectangular projection (ERP) panoramic images for 3D detection. It introduces the ERP-Ray Geometry Adapter (ERGA-Ray) to model spherical rays, the Visual-Action Reasoning Chain (VARC) to ground hypotheses in panoramic evidence, and the Appearance-Guided Heading Expert (AGHE) to recover high‑resolution heading cues. The method achieves a 2.96 NDS improvement on Spheriverse and a 24.87 mAP gain on PanoMMOcc, while preserving most of the baseline’s performance across configurations.
By Runtong Wu, Fei Teng, Di Wen, Guoqiang Zhao, Kunyu Peng, Kailun Yang
arXiv:2607.05568v2 Announce Type: replace-cross
Abstract: Compact primitive abstractions represent 3D shapes with a few geometric primitives while preserving recognizable components. Learned methods...
By Gregor Kobsik, Tim Elsner, Leif Kobbelt
arXiv:2607. 02561v1 Announce Type: cross Abstract: Consumer depth sensors such as the LiDAR scanner on recent iPhones provide metric range, but their useful range is short and their returns are sparse.
By Jinwen Wen
arXiv:2607. 09825v1 Announce Type: cross Abstract: Robotic manipulation policies rely on pre-trained vision models that give either a global scene embedding or a dense patch grid.
By Yi Li (TU Darmstadt), Alexandre Chapin (LIRIS), Liming Chen (LIRIS), Jan Peters (TU Darmstadt), Alap Kshirsagar (IIT Delhi, ADU)
arXiv:2608. 19536v1 Announce Type: cross Abstract: Learning-based global point cloud registration has achieved remarkable progress, yet its reliance on geometric representations makes existing methods sensitive to variations in point density, scan pattern, viewpoint, and sensor characteristics.
By Eunsoo Im, Junghun Suh, Gyeonggwan Lee, Seunghwan Hong