arXivDaily arXiv每日学术速递 周一至周五更新

视觉与机器人

3D 视觉

三维重建、NeRF、Gaussian Splatting、点云和空间智能。

2026-06-18 至 2026-06-18 共收录 11 信号源:cs.CV, cs.GR, cs.RO

1. 点云 11 篇

2605.17131 2026-06-18 cs.CV cs.AI cs.LG 版本更新 84%

A Survey on Deep Learning Architectures for Point Cloud Classification and Segmentation

针对点云分类和分割的深度学习架构系统性调研

Minhas Kamal, Hiranya Garbha Kumar, Balakrishnan Prabhakaran

机构 * State University of New York at Albany(纽约州立大学阿尔巴尼分校)

专题命中 点云 :point cloud(title,abstract);3D vision(abstract);分类 cs.CV

AI总结 本文系统性地探讨了点云分类和分割中的深度学习架构,分析了点云数据的结构特性,分类了不同架构的工作,并评估了其在主流基准上的性能,同时指出了开放挑战和未来方向。

Comments We reviewed a decade of advancements in point cloud processing: trace the evolution of the field from its foundational roots to the modern SOTA, analyze how diverse architectures overcome the inherent geometric challenges of 3D data, and map out critical research gaps alongside promising future directions. GitHub: https://github.com/MinhasKamal/DeepLearningForPointCloud

Journal ref ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.18472 2026-06-18 cs.CV 新提交 83%

Domain Generalizable Adaptation of 3D Vision-Language Models via Regularized Fine-Tuning

通过正则化微调实现可域泛化的3D视觉-语言模型适应

Sneha Paul, Zachary Patterson, Nizar Bouguila

机构 * Concordia University(康考迪亚大学)

专题命中 点云 :3D vision(title,abstract);point cloud(abstract);分类 cs.CV

AI总结 提出ReFine3D框架,通过选择性层调优、多视图一致性、同义词提示及点渲染视觉监督等正则化策略,提升3D大语言模型在域泛化中的性能。

Comments Accepted at Transactions on Machine Learning Research (TMLR)

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.18787 2026-06-18 cs.CV 新提交 79%

Learned Radius Estimation for UDF-Based Point Cloud Reconstruction

基于UDF的点云重建中的学习半径估计

Eito Ogawa, Hiroshi Watanabe

机构 * Graduate School of FSE Waseda University Tokyo, Japan(Waseda大学研究生院FSE学院东京日本)

专题命中 点云 :point cloud(title,abstract);分类 cs.CV

AI总结 提出一种学习型逐查询半径选择器,预测连续支撑半径并插入冻结的LoSF-UDF骨干网络,通过抛物线插值获取离网目标半径进行训练,提高点云表面重建的细粒度精度。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.01200 2026-06-18 cs.CV eess.IV 版本更新 79%

Objective Quality Assessment of Point Clouds Using Multi-scale Implicit Structural Similarity

点云的多尺度隐式结构相似性客观质量评估

Zhang Chen, Shuai Wan, Yuezhe Zhang, Siyu Ren, Fuzheng Yang, Junhui Hou

机构 * School of Electronics and Information, Northwestern Polytechnical University(电子与信息学院,西北工业大学) Department of Computer Science, City University of Hong Kong(计算机科学系,香港城市大学) School of Telecommunication Engineering, Xidian University(电信工程学院,西安电子科技大学)

专题命中 点云 :point cloud(title,abstract);分类 cs.CV

AI总结 针对点云质量评估中不规则数据匹配困难的问题,提出多尺度隐式结构相似性度量(MS-ISSM),通过径向基函数连续表示局部特征并比较隐式函数系数,结合ResGrouped-MLP网络,在多个基准上超越现有方法。

Comments IEEE TMM Accepted

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.18583 2026-06-18 cs.CV cs.RO 新提交 62%

Aerial-ground LiDAR place recognition with patch-level self-supervised learning and expanded reciprocal re-ranking

空地激光雷达地点识别:基于块级自监督学习和扩展互逆重排序

Yandi Yang, Xianghong Zou, Jianping Li, Haofeng Xie, Saurav Uprety, Hongzhou Yang, Naser El-Sheimy

机构 * University of Calgary(卡尔加里大学) Nanchang University(南昌大学) Nanyang Technological University(南洋理工大学) Wuhan University(武汉大学)

专题命中 点云 :point cloud(abstract);分类 cs.CV、cs.RO

AI总结 提出一种空地激光雷达地点识别框架,通过多尺度块级自监督学习缩小域差距,并利用扩展互逆重排序算法减少误检,在多个数据集上显著提升检索精度。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.19154 2026-06-18 cs.RO 新提交 57%

Viking Hill Dataset: A Lidar-Radar-Camera Dataset for Detection and Segmentation in Forest Scenes

Viking Hill数据集:用于森林场景检测与分割的激光雷达-雷达-相机数据集

Vladimír Kubelka, Oleksandr Kotlyar, Unal Artan, Martin Magnusson

机构 * Örebro University(奥雷布罗大学) AASS research centre(AASS研究中心) Robot Navigation and Perception Lab(机器人导航与感知实验室)

专题命中 点云 :point cloud(abstract);分类 cs.RO

AI总结 提出首个包含4D成像雷达的森林多传感器数据集,通过MinkowskiUNet实现雷达与激光雷达点云的语义分割,并评估树干分割质量与树木尺寸的关系。

Comments 33 pages, 11 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.18952 2026-06-18 cs.CV 新提交 57%

SP-TransientBench: A Real-Captured Single Photon Perception Benchmark

SP-TransientBench: 一个真实捕获的单光子感知基准

Hongzhou Dong, Zili Zhang, Ziting Wen, Yiheng Qiang, Runrong Deng, Wenle Dong, Ziwen Jiang, Xinyang Li, Rui Lu, Shuoyao Sun, Wenyu Wang, Ziyi Xia, Haitao Zheng, Guodong Shi, Xiaoqiang Ren

机构 * Shanghai University(上海大学) Southern University of Science and Technology(南方科技大学) The University of Sydney(悉尼大学)

专题命中 点云 :3D vision(abstract);分类 cs.CV

AI总结 针对单光子LiDAR在真实场景中因噪声和多回波瞬态现象导致的感知挑战,提出包含10个场景、10297个视角的真实捕获多任务基准STB,支持深度估计、多视图重建和3D语义理解评估。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.18948 2026-06-18 cs.RO 新提交 57%

C-ARC: Continuous-Adaptive Range Clustering for Non-Repetitive LiDAR Sensors

C-ARC: 面向非重复式LiDAR传感器的连续自适应范围聚类

Nick B. Schroeder, Jonathan Lichtenfeld, Oskar von Stryk

机构 * Technical University of Darmstadt(德累斯顿技术大学) Simulation, Systems Optimization and Robotics Group(仿真、系统优化与机器人组)

专题命中 点云 :point cloud(abstract);分类 cs.RO

AI总结 提出C-ARC框架,通过滑动窗口上的持久双图结构解耦高频点插入与按需聚类检索,并利用指数控制环自适应校准网格分辨率,实现非重复式LiDAR点云的实时聚类。

Comments Submitted to IEEE Robotics and Automation Letters. This work has been submitted to the IEEE for possible publication. 8 pages, 7 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.08206 2026-06-18 cs.CV cs.LG 新提交 57%

SegmentAnyTreeV2: Scaling Transformer-Based Tree Instance Segmentation Across Sensors, Platforms, and Forests

SegmentAnyTreeV2:跨传感器、平台和森林的基于Transformer的树木实例分割扩展

Maciej Wielgosz, Stefano Puliti, Rasmus Astrup

机构 * Norwegian Institute of Bioeconomy Research (NIBIO)(挪威生物经济研究所(NIBIO))

专题命中 点云 :point cloud(abstract);分类 cs.CV

AI总结 提出SegmentAnyTreeV2,一种传感器和平台无关的森林点云语义与实例分割框架,结合Point Transformer v3骨干网络、轻量语义头和树木交叉注意力掩码解码器,在FOR-instance v3基准上达到90.5%精度和80.2%召回率,并展现出强跨域泛化能力。

Comments 25 pages, 6 figures, 10 tables, Corrected bibliography metadata and minor typographical issues; results unchanged

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.21583 2026-06-18 cs.CV 版本更新 57%

HACMatch Semi-Supervised Rotation Regression with Hardness-Aware Curriculum Pseudo Labeling

HACMatch: 基于难度感知课程伪标签的半监督旋转回归

Mei Li, Huayi Zhou, Suizhi Huang, Yuxiang Lu, Yue Ding, Hongtao Lu

机构 * Shanghai Jiao Tong University(上海交通大学)

专题命中 点云 :point cloud(abstract);分类 cs.CV

AI总结 提出一种难度感知课程学习框架,通过动态选择伪标签样本和结构化数据增强,在少量标注数据下提升半监督旋转回归性能。

Comments This is an accepted manuscript of an article published in Computer Vision and Image Understanding

Journal ref Computer Vision and Image Understanding (2026)

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.19228 2026-06-18 astro-ph.EP astro-ph.SR 新提交 50%

JWST-TST High Contrast: First Direct Spectroscopy of GJ 504 b reveals Clouds and Possible Metal Enrichment

JWST-TST 高对比度:GJ 504 b 的首次直接光谱揭示云和可能的金属富集

Aneesh Baburaj, Jean-Baptiste Ruffio, Marshall Perrin, Jerry W. Xuan, William O. Balmer, Yayaati Chachan, Quinn M. Konopacky, Travis S. Barman, Mathilde Mâlin, Kielan K. W. Hoch, Emily Rickman, Kimberly Ward-Duong, Laurent Pueyo, Julien H. Girard, Isabel Rebollido, Alexis Bidot, Christine Chen, Kadin Worthen, Cicero Lu, Jens Kammerer, Roeland P. van der Marel, Nikole K. Lewis, Jeff Valenti, Sara Seager, Chris Stark, Rémi Soummer, Jay Anderson, Charles-Philippe Lajoie, Mark Clampin, C. Matt Mountain

专题命中 点云 :point cloud(abstract)

AI总结 利用 JWST/NIRSpec 对直接成像行星质量伴星 GJ 504 b 进行中分辨率光谱观测,通过先进后处理技术检测到强信号,提取 2.9-5.3 μm 光谱并建模,发现多种分子、非平衡化学和盐云,推断质量约 25.2 M_Jup,金属丰度高于主星,支持行星形成机制。

Comments 35 pages, 20 figures, 6 tables

Journal ref AJ 172 28 (2026)

详情

展开后加载摘要…

URL PDF HTML 收藏