检索结果-内蒙古大学图书馆

Weakly Supervised Segmentation on Outdoor 4d Point Clouds With Progressive 4d Grouping

IEEE TRANSACTIONS ON PATTERN ANALYSIS ANd MACHINE INTELLIGENCE 2025年第5期47卷 3487-3499页

作者： Shi, Hanyu Liu, Fayao Wu, Zhonghua Xu, Yi Lin, Guosheng Nanyang Technol Univ Singapore 639798 Singapore ASTAR Inst Infocomm Res Singapore 138632 Singapore GoerTek Elect Inc Santa Clara CA 95054 USA

Recently, some weakly supervised 3d point cloud segmentation methods have been proposed to develop effective models with minimum annotation efforts. Our previous work, W4dTS, proposes a challenging task that utilizes only 0.001% points in outdoor point cloud datasets to achieve an effective segmentation model. However, under an extremely limited annotation budget, the quality of pseudo labels generated by W4dTS is unsatisfactory, which limits the segmentation performance in such scenarios. To solve this issue, we propose a progressive 4d grouping approach to group the annotated and unannotated points across space and time, which can generate high-quality pseudo labels with very sparse annotated points. Moreover, to further improve our progressive 4d grouping approach, we design a cross-frame contrastive learning and a local consistency learning to improve the quality of our 4d grouping. Experimental results reveal that with only 0.001% annotations, our solution significantly outperforms the previous best approach on SemanticKITTI. We also evaluate our framework on the SemanticPOSS dataset and ScribbleKITTI dataset, and achieve performances close to our fully supervised backbone models.

关键词： 3d computer vision semantic segmentation weakly supervised semantic segmentation 3d computer vision semantic segmentation weakly supervised semantic segmentation

来源：评论

学校读者我要写书评

暂无评论

Near Real-Time 3d Reconstruction of Construction Sites Based on Surveillance Cameras

引用

BUILdINGS 2025年第4期15卷 567-567页

作者： Sun, Aoran An, Xuehui Li, Pengfei Lv, Miao Liu, Wenzhe Tsinghua Univ State Key Lab Hydrosci & Engn Beijing 100084 Peoples R China Tsinghua Univ Dept Hydraul Engn Beijing 100084 Peoples R China Chongqing Jiaotong Univ Engn Res Ctr Diag Technol Hydroconstruct Chongqing 400074 Peoples R China

The 3d reconstruction of construction sites is of great importance for construction progress, quality, and safety management. Currently, most of the existing 3d reconstruction methods are unable to conduct continuous and uninterrupted perception, and it is difficult to achieve registration with real coordinates and dimensions. This study proposes a hierarchical registration framework for 3d reconstruction of construction sites based on surveillance cameras. This method can quickly perform on-site 3d reconstruction and restoration by taking surveillance camera images as inputs. It combines 2d and 3d features and does not need transfer learning or camera calibration. By experimenting on one construction site, we found that this framework can complete the 3d point cloud estimation and registration of construction sites within an average of 3.105 s through surveillance images. The average RMSE of the point cloud within the site is 0.358 m, which is better than most point cloud registration methods. Through this method, 3d data within the scope of surveillance cameras can be quickly obtained, and the connection between 2d and 3d can be effectively established. Combined with visual information, it is beneficial to the digital twin management of construction sites.

关键词： image-based 3d reconstruction 3d computer vision surveillance camera Structure from Motion (SfM) point cloud registration

来源：评论

学校读者我要写书评

暂无评论

GTIGNet: Global Topology Interaction Graphormer Network for 3d hand pose estimation

引用

NEURAL NETWORKS 2025年 185卷 107221页

作者： Liu, Yanjun Fan, Wanshu Wang, Cong Wen, Shixi Yang, Xin Zhang, Qiang Wei, Xiaopeng Zhou, dongsheng Dalian Univ Sch Software Engn Natl & Local Joint Engn Lab Comp Aided Design Dalian Peoples R China Ctr Adv Reliabil & Safety CAiRS Hong Kong Peoples R China Dalian Univ Sch Informat Engn Dalian Peoples R China Dalian Univ Technol Sch Comp Sci & Technol Dalian Peoples R China

Estimating 3d hand poses from monocular RGB images presents a series of challenges, including complex hand structures, self-occlusions, and depth ambiguities. Existing methods often fall short of capturing the longdistance dependencies of skeletal and non-skeletal connections for hand joints. To address these limitations, we introduce the Global Topology Interaction Graphormer Network (GTIGNet), a novel deep learning architecture designed to improve 3d hand pose estimation. Our model incorporates a Context-Aware Attention Block (CAAB) within the 2d pose estimator to enhance the extraction of multi-scale features, yielding more accurate 2d joint heatmaps to support the task that followed. Additionally, we introduce a High-Order Graphormer that explicitly and implicitly models the topological structure of hand joints, thereby enhancing feature interaction. Ablation studies confirm the effectiveness of our approach, and experimental results on four challenging datasets, Rendered Hand dataset (RHd), Stereo Hand Pose Benchmark (STB), First-Person Hand Action Benchmark (FPHA), and FreiHANd dataset, indicate that GTIGNet achieves state-of-the-art performance in 3d hand pose estimation. Notably, our model achieves an impressive Mean Per Joint Position Error (MPJPE) of 9.98 mm on RHd, 6.12 mm on STB, 11.15 mm on FPHA and 10.97 mm on FreiHANd.

关键词： 3d hand pose estimation Transformer GCN Topology 3d computer vision

来源：评论

学校读者我要写书评

暂无评论

3d depthwise Convolution: Reducing Model Parameters in 3d vision Tasks 32nd

3D Depthwise Convolution: Reducing Model Parameters in 3D Vi...

引用

32nd Canadian Conference on Artificial Intelligence (Canadian AI)

作者： Ye, Rongtian Liu, Fangyu Zhang, Liqiang Beijing Normal Univ Beijing Peoples R China Univ Waterloo Waterloo ON Canada

ISBN: (纸本)9783030183059;9783030183042

Standard 3d convolution operations usually require larger amounts of memory and computation cost than 2d convolution operations. The fact increases the difficulty of the development of deep neural nets in many 3d vision tasks. In this paper, we investigate the possibility of applying depthwise separable convolutions in 3d scenario and introduce the use of 3d depthwise convolution. A 3d depthwise convolution splits a single standard 3d convolution into two separate steps, which would drastically reduce the number of parameters in 3d convolutions with more than one order of magnitude. We experiment with 3d depthwise convolution on popular CNN architectures and also compare it with a similar structure called pseudo-3d convolution. The results demonstrate that, with 3d depthwise convolutions, 3d vision tasks like classification and reconstruction can be carried out with more light-weighted neural networks while still delivering comparable performances.

关键词： depthwise convolution Low latency models Pseudo-3d convolution 3d computer vision 3d Reconstruction

来源：评论

学校读者我要写书评

暂无评论

PCT:Point cloud transformer

引用

Computational Visual Media 2021年第2期7卷 187-199页

作者： Meng-Hao Guo Jun-Xiong Cai Zheng-Ning Liu Tai-Jiang Mu Ralph R.Martin Shi-Min Hu BNRist Department of Computer Science and TechnologyTsinghua UniversityBeiing 100084China Cardiff University Cardiff CF243

The irregular domain and lack of ordering make it challenging to design deep neural networks for point cloud *** paper presents a novel framework named Point Cloud Transformer(PCT)for point cloud *** is based on Transformer,which achieves huge success in natural language processing and displays great potential in image *** is inherently permutation invariant for processing a sequence of points,making it well-suited for point cloud *** better capture local context within the point cloud,we enhance input embedding with the support of farthest point sampling and nearest neighbor *** experiments demonstrate that the PCT achieves the state-of-the-art performance on shape classification,part segmentation,semantic segmentation,and normal estimation tasks.

关键词： 3d computer vision deep learning point cloud processing Transformer

来源：评论

学校读者我要写书评

暂无评论

A Survey of Robustness and Safety of 2d and 3d deep Learning Models against Adversarial Attacks

引用

ACM COMPUTING SURVEYS 2024年第6期56卷 1-37页

作者： Li, Yanjie Xie, Bin Guo, Songtao Yang, Yuanyuan Xiao, Bin Hong Kong Polytech Univ Hung Hom KLN 11 Yuk Choi Rd Hong Kong Peoples R China Chongqing Univ 174 Shapingba Zheng St Chongqing Peoples R China SUNY Stony Brook New Comp Sci Engn Dr Stony Brook NY 11794 USA

Benefiting from the rapid development of deep learning, 2d and 3d computer vision applications are deployed in many safe-critical systems, such as autopilot and identity authentication. However, deep learning models are not trustworthy enough because of their limited robustness against adversarial attacks. The physically realizable adversarial attacks further pose fatal threats to the application and human safety. Lots of papers have emerged to investigate the robustness and safety of deep learning models against adversarial attacks. To lead to trustworthy AI, we first construct a general threat model from different perspectives and then comprehensively review the latest progress of both 2d and 3d adversarial attacks. We extend the concept of adversarial examples beyond imperceptive perturbations and collate over 170 papers to give an overview of deep learning model robustness against various adversarial attacks. To the best of our knowledge, we are the first to systematically investigate adversarial attacks for 3d models, a flourishing field applied to many real-world applications. In addition, we examine physical adversarial attacks that lead to safety violations. Last but not least, we summarize present popular topics, give insights on challenges, and shed light on future research on trustworthy AI.

关键词： deep learning 3d computer vision adversarial attack robustness

来源：评论

学校读者我要写书评

暂无评论

An overview of depth cameras and range scanners based on time-of-flight technologies

引用

MACHINE vision ANd APPLICATIONS 2016年第7期27卷 1005-1020页

作者： Horaud, Radu Hansard, Miles Evangelidis, Georgios Menier, Clement INRIA Grenoble Rhone Alpes Res Montbonnot St Martin France INRIA Grenoble Rhone Alpes Montbonnot St Martin France Queen Mary Univ London Sch Elect Engn & Comp Sci London England 4D View Solut Grenoble France

Time-of-flight (TOF) cameras are sensors that can measure the depths of scene points, by illuminating the scene with a controlled laser or LEd source and then analyzing the reflected light. In this paper, we will first describe the underlying measurement principles of time-of-flight cameras, including: (1) pulsed-light cameras, which measure directly the time taken for a light pulse to travel from the device to the object and back again, and (2) continuous-wave-modulated light cameras, which measure the phase difference between the emitted and received signals, and hence obtain the travel time indirectly. We review the main existing designs, including prototypes as well as commercially available devices. We also review the relevant camera calibration principles, and how they are applied to TOF devices. Finally, we discuss the benefits and challenges of combined TOF and color camera systems.

关键词： LIdAR Range scanners Single-photon avalanche diode Time-of-flight cameras 3d computer vision Active light sensors

来源：评论

学校读者我要写书评

暂无评论

BushNet: Effective semantic segmentation of bush in large-scale point clouds

引用

computerS ANd ELECTRONICS IN AGRICULTURE 2022年 193卷 106653-106653页

作者： Wei, Hejun Xu, Enyong Zhang, Jinlai Meng, Yanmei Wei, Jin dong, Zhen Li, Zhengqiang Guangxi Univ Coll Mech Engn Naning 530004 Guangxi Peoples R China Dongfeng Liuzhou Motor Co Ltd Liuzhou 545005 Guangxi Peoples R China Huazhong Univ Sci & Technol Coll Mech Sci & Engn Wuhan 430074 Hubei Peoples R China

Effective and robust semantic segmentation of bush is the fundamental problem of agroforestry environment understanding. However, the point cloud data of most large-scale agroforestry scenes is extremely large, and it is difficult to perform semantic segmentation on them. In order to realize the effective semantic segmentation of bush point cloud in large-scale agroforestry environment, this paper proposes BushNet, a novel point cloud segmentation network consists of three key components. Firstly, we propose the minimum probability random sampling module which can quickly and randomly sample a huge point cloud while avoiding the problem of random sampling easily causing re-sampling, reducing the consumption of computing resources and improving the convergence speed. Secondly, we propose the local multi-dimensional feature fusion module which makes the network more sensitive to bush point cloud features, thereby showing better bush segmentation performance. Thirdly, we propose the multi-channel attention module to achieve more accurate attention distribution and improved training efficiency. Experiments demonstrate that our approach significantly improves segmentation performance on multiple large-scale agroforestry point cloud data sets.

关键词： deep learning 3d computer vision Point cloud

来源：评论

学校读者我要写书评

暂无评论

Multi-sensor 3d object dataset for object recognition with full pose estimation

引用

NEURAL COMPUTING & APPLICATIONS 2017年第5期28卷 941-952页

作者： Garcia-Garcia, Alberto Orts-Escolano, Sergio Oprea, Sergiu Garcia-Rodriguez, Jose Azorin-Lopez, Jorge Saval-Calvo, Marcelo Cazorla, Miguel Univ Alicante Comp Technol Dept Alicante Spain Univ Alicante Comp Sci & Artificial Intelligence Dept Alicante Spain

In this work, we propose a new dataset for 3d object recognition using the new high-resolution Kinect V2 sensor and some other popular low-cost devices like PrimeSense Carmine. Since most already existing datasets for 3d object recognition lack some features such as 3d pose information about objects in the scene, per pixel segmentation or level of occlusion, we propose a new one combining all this information in a single dataset that can be used to validate existing and new 3d object recognition algorithms. Moreover, with the advent of the new Kinect V2 sensor we are able to provide high-resolution data for RGB and depth information using a single sensor, whereas other datasets had to combine multiple sensors. In addition, we will also provide semiautomatic segmentation and semantic labels about the different parts of the objects so that the dataset could be used for testing robot grasping and scene labeling systems as well as for object recognition.

关键词： 3d computer vision Object recognition 3d object dataset Kinect V2 PrimeSense Carmine

来源：评论

学校读者我要写书评

暂无评论

A Performance Evaluation of Correspondence Grouping Methods for 3d Rigid data Matching

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS ANd MACHINE INTELLIGENCE 2021年第6期43卷 1859-1874页

作者： Yang, Jiaqi Xian, Ke Wang, Peng Zhang, Yanning Northwestern Polytech Univ Natl Engn Lab Integrated Aerosp Ground Ocean Big Sch Comp Sci Xian 710129 Peoples R China Huazhong Univ Sci & Technol Natl Key Lab Sci & Technol Multispectral Informat Sch Artificial Intelligence & Automat Wuhan 430074 Peoples R China

Seeking consistent point-to-point correspondences between 3d rigid data (point clouds, meshes, or depth maps) is a fundamental problem in 3d computer vision. While a number of correspondence selection methods have been proposed in recent years, their advantages and shortcomings remain unclear regarding different applications and perturbations. To fill this gap, this paper gives a comprehensive evaluation of nine state-of-the-art 3d correspondence grouping methods. A good correspondence grouping algorithm is expected to retrieve as many as inliers from initial feature matches, giving a rise in both precision and recall as well as facilitating accurate transformation estimation. Toward this rule, we deploy experiments on three benchmarks with different application contexts, including shape retrieval, 3d object recognition, and point cloud registration. We also investigate various perturbations such as noise, point density variation, clutter, occlusion, partial overlap, different scales of initial correspondences, and different combinations of keypoint detectors and descriptors. The rich variety of application scenarios and nuisances result in different spatial distributions and inlier ratios of initial feature correspondences, thus enabling a thorough evaluation. Based on the outcomes, we give a summary of the traits, merits, and demerits of evaluated approaches and indicate some potential future research directions.

关键词： Three-dimensional displays Shape Measurement detectors Feature extraction Object recognition Clutter Performance evaluation correspondence grouping 3d computer vision 3d rigid data shape matching

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：