检索结果-内蒙古大学图书馆

Robust detection, segmentation, and Metrology of High Bandwidth Memory 3d Scans Using an Improved Semi-Supervised deep Learning Approach

引用

SENSORS 2023年第12期23卷 5470-5470页

作者： Wang, Jie Chang, Richard Zhao, Ziyuan Pahwa, Ramanpreet Singh ASTAR Inst Infocomm Res I2R 1 Fusionopolis Way21-01Connexis South Tower Singapore 138632 Singapore

Recent advancements in 3d deep learning have led to significant progress in improving accuracy and reducing processing time, with applications spanning various domains such as medical imaging, robotics, and autonomous vehicle navigation for identifying and segmenting different structures. In this study, we employ the latest developments in 3d semi-supervised learning to create cutting-edge models for the 3d object detection and segmentation of buried structures in high-resolution X-ray semiconductors scans. We illustrate our approach to locating the region of interest of the structures, their individual components, and their void defects. We showcase how semi-supervised learning is utilized to capitalize on the vast amounts of available unlabeled data to enhance both detection and segmentation performance. Additionally, we explore the benefit of contrastive learning in the data pre-selection step for our detection model and multi-scale Mean Teacher training paradigm in 3d semantic segmentation to achieve better performance compared with the state of the art. Our extensive experiments have shown that our method achieves competitive performance and is able to outperform by up to 16% on object detection and 7.8% on semantic segmentation. Additionally, our automated metrology package shows a mean error of less than 2 mu m for key features such as Bond Line Thickness and pad misalignment.

关键词： 3d semi-supervised Learning 3d object detection 3d semantic segmentation contrastive learning 3d metrology

来源：评论

学校读者我要写书评

暂无评论

defect detection and repair algorithm for structures generated by topology optimization based on 3d hierarchical fully convolutional network

引用

Biomimetic Intelligence & Robotics 2024年第2期4卷 24-31页

作者： Zhiyu Wan Hai Lan Sichao Lin Houde dai School of Advanced Manufacturing Fuzhou UniversityJinjiang 362251China Quanzhou Institute of Equipment Manufacturing Haixi InstitutesChinese Academy of SciencesJinjiang 362216China Fujian Key Laboratory of Special Intelligent Equipment Safety Measurement and Control Fujian Special Equipment Inspection and Research InstituteFuzhou 350008China

Customized 3d-printed structural parts are widely used in surgical *** satisfy the mechanical properties and kinematic functions of these structural parts,a topology optimization technique is adopted to obtain the optimal structural layout while meeting the constraints and ***,topology optimization currently faces some practical challenges that must be addressed,such as ensuring that structures do not have significant defects when converted to additive manufacturing *** address this problem,we designed a 3d hierarchical fully convolutional network(FCN)to predict the precise position of the defective *** on the prediction results,an effective repair strategy is adopted to repair the defective structure.A series of experiments is conducted to demonstrate the effectiveness of our *** to the 2d fully convolutional network and the rule-based detection method,our approach can accurately capture most defect structures and achieve 89.88%precision and 95.59%***,we investigate the impact of different ways to increase the receptive field of our model,as well as the trade-off between different defect-repairing *** results of the experiment demonstrate that the hierarchical structure,which increases the receptive field,can substantially improve the defect detection *** the best of our knowledge,this paper is the first to investigate 3d defect prediction and repair for topology optimization in conjunction with deep learning algorithms,providing practical tools and new perspectives for the subsequent development of topology optimization techniques.

关键词： Topology optimization Additive manufacturing deep learning 3d semantic segmentation defect detection

来源：评论

学校读者我要写书评

暂无评论

segmentation OF PLANT POINT CLOUdS FOR PHENOTYPING 31

SEGMENTATION OF PLANT POINT CLOUDS FOR PHENOTYPING

引用

31st IEEE Conference on Signal Processing and Communications Applications (SIU)

作者： Yavuz, Ahmetcan Arda, Berke dutagaci, Helin Sankur, Bulent Bogazici Univ Elekt Elekt Muhendisli Istanbul Turkiye Eskisehir Osmangazi Univ Elekt Elekt Muhendisli Eskisehir Turkiye

ISBN: (纸本)9798350343557

In this paper, we consider semantic segmentation of 3d plant point clouds. We propose various modifications to improve the performance of two main state-of-the-art methods for semantic segmentation of plants. In the first method, we consider the classical machine learning approach where local geometric features are classified via SVM. We propose to process the point cloud in two stages: The points classified, in the first stage, with high confidence are removed to enable retraining of the SVM in the second stage. Furthermore, we apply a multi-resolution approach to feature extraction. The second state-of-the-art method is PointNet++, which is a widely used deep learning architecture for plant segmentation. We propose alternative approaches, based on critical points and guide points, for data preprocessing for PointNet++. We observed that these modifications improved the segmentation performance on the tomato models of the Pheno4d data set.

关键词： plant segmentation 3d semantic segmentation 3d point cloud PointNet plus

来源：评论

学校读者我要写书评

暂无评论

PLANesT-3d: A new annotated data set of 3d color point clouds of plants 31

PLANesT-3D: A new annotated data set of 3D color point cloud...

引用

31st IEEE Conference on Signal Processing and Communications Applications (SIU)

作者： Mertoglu, Kerem Salk, Yusuf Sarikaya, Server Karahan Turgut, Kaya Evrenosoglu, Yasemin Cevikalp, Hakan Gerek, Omer Nezih dutagaci, Helin Rousseau, david Usak Univ Ziraat Fak Usak Turkiye Eskisehir Osmangazi Univ Elekt Elekt Muhendisligi Eskisehir Turkiye Eskisehir Osmangazi Univ Ziraat Fak Eskisehir Turkiye Eskisehir Tekn Univ Elekt Elekt Muhendisligi Eskisehir Turkiye Univ Angers LARIS Angers France

ISBN: (纸本)9798350343557

In this paper, we introduce a new data set, named as PLANesT-3d, which includes complete 3d color point clouds of plants. PLANesT-3d is composed of 34 point clouds acquired from three different plant species: Pepper, rosebush and ribes. The point clouds were reconstructed from 2d color images of real plants through multiview stereo. Through a semi-automatic process, background and noise removal was performed. The point clouds were scaled to their true metric dimensions and brought to a particular pose. Each point in the point clouds was manually labeled into "leaf" and "stem" classes. Also, each leaf instance was assigned a leaf identification number. In order to provide benchmark semantic segmentation performance, segmentation results obtained with the application of PointNet++ are reported. We believe that PLANesT-3d data set will contribute to the development and assessment of segmentation methods that operate on 3d color plant point clouds.

关键词： 3d point cloud 3d plant modeling 3d plant dataset 3d semantic segmentation

来源：评论

学校读者我要写书评

暂无评论

A Novel Approach Based on Cluster-group for Classification of 3d Residential Scene

A Novel Approach Based on Cluster-group for Classification o...

引用

International Conference on Information Science, Electronics and Electrical Engineering (ISEEE)

作者： Lu, Guiliang Zhou, Yu Yu, Yao du, Sidan Nanjing Univ Sch Elect Sci & Engn Nanjing Peoples R China

ISBN: (纸本)9781479931972

To understand scenes and help autonomous robots and cars, researchers' attention is directed through the problem of classifying 3d point cloud. In this paper, we present a novel approach to semantically segment 3d point cloud of residential scenes captured by a lidar sensor. Our approach is based on a dual-scale analysis: a small-scale clustering and a large-scale grouping. Features used to train our AdaBoost classifier are then extracted from clusters and groups. We evaluate our method with a challenging lidar data set. The result shows our approach can classify scene objects accurately.

关键词： 3d semantic segmentation Classification Point Cloud Lidar

来源：评论

学校读者我要写书评

暂无评论

Enhancing Autonomous Train Safety Through A Priori-Map Based Perception 4th

Enhancing Autonomous Train Safety Through A Priori-Map Based...

引用

4th International Conference on Reliability, Safety, and Security of Railway Systems (RSSRail)

作者： Mahtani, Ankur Chouchani, Nadia Herbreteau, Maxime Rafin, denis FCS Railenium 180 Rue Joseph Louis Lagrange F-59300 Famars France Thales Serv Numer SAS 290 Allee Lac F-31670 Labege France SpirOps 8 Passage Bonne Graine F-75011 Paris France

ISBN: (纸本)9783031058134;9783031058141

Autonomous driving tends to increase use of perception as a tool for analyzing the environment before making a decision that could impact driving. However, recent techniques based on machine learning do not provide the necessary interpretability to ensure sufficient driving safety. Combining multiple sources, deterministic or not, allows results to be cross-referenced and therefore more reliable. In this paper, we propose a novel methodology that aligns an infrastructure mapping system and point cloud analysis for railway tracks and catenaries perception to ensure autonomous train's safety. By using a deep learning model to recognize and classify rails with the implicit knowledge of the railway infrastructure, we exceed in performance all previous systems of infrastructure: 60.9% in mIoU for tracks segmentation and 9.27 points mMink for points alignment with ground-truth, at an interesting runtime of 20 Hz. Moreover, we propose an embedded solution for automatic monitoring which avoids hours of maintenance traffic on the railway tracks. This solution is used as acquisition system feeding map and perception in real-world data for autonomous trains.

关键词： Autonomous train Railway map Point cloud 3d semantic segmentation deep learning

来源：评论

学校读者我要写书评

暂无评论

GOV-NeSF: Generalizable Open-Vocabulary Neural semantic Fields

GOV-NeSF: Generalizable Open-Vocabulary Neural Semantic Fiel...

引用

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Wang, Yunsong Chen, Hanlin Lee, Gim Hee Natl Univ Singapore Dept Comp Sci Singapore Singapore

ISBN: (纸本)9798350353006

Recent advancements in vision-language foundation models have significantly enhanced open-vocabulary 3d scene understanding. However, the generalizability of existing methods is constrained due to their framework designs and their reliance on 3d data. We address this limitation by introducing Generalizable Open-Vocabulary Neural semantic Fields (GOV-NeSF), a novel approach offering a generalizable implicit representation of 3d scenes with open-vocabulary semantics. We aggregate the geometry-aware features using a cost volume, and propose a Multi-view Joint Fusion module to aggregate multi-view features through a cross-view attention mechanism, which effectively predicts view-specific blending weights for both colors and open-vocabulary features. Remarkably, our GOV-NeSF exhibits state-of-the-art performance in both 2d and 3d open-vocabulary semantic segmentation, eliminating the need for ground truth semantic labels or depth priors, and effectively generalize across scenes and datasets without fine-tuning.

关键词： 3d Scene Understanding 3d semantic segmentation Neural Radiance Fields Open-Vocabulary

来源：评论

学校读者我要写书评

暂无评论

RANP: Resource Aware Neuron Pruning at Initialization for 3d CNNs 8

RANP: Resource Aware Neuron Pruning at Initialization for 3D...

引用

8th International Conference on 3d Vision (3dV)

作者： Xu, Zhiwei Ajanthan, Thalaiyasingam Vineet, Vibhav Hartley, Richard Australian Natl Univ Canberra ACT Australia Australian Ctr Robot Vis Canberra ACT Australia Microsoft Res Redmond WA USA CSIRO Data61 Brisbane Qld Australia

ISBN: (纸本)9781728181288

Although 3d Convolutional Neural Networks (CNNs) are essential for most learning based applications involving dense 3d data, their applicability is limited due to excessive memory and computational requirements. Compressing such networks by pruning therefore becomes highly desirable. However, pruning 3d CNNs is largely unexplored possibly because of the complex nature of typical pruning algorithms that embeds pruning into an iterative optimization paradigm. In this work, we introduce a Resource Aware Neuron Pruning (RANP) algorithm that prunes 3d CNNs at initialization to high sparsity levels. Specifically, the core idea is to obtain an importance score for each neuron based on their sensitivity to the loss function. This neuron importance is then reweighted according to the neuron resource consumption related to FLOPs or memory. We demonstrate the effectiveness of our pruning method on 3d semantic segmentation with widely used 3d-UNets on ShapeNet and BraTS'18 as well as on video classification with MobileNetV2 and I3d on UCF101 dataset. In these experiments, our RANP leads to roughly 50%-95% reduction in FLOPs and 35%-80% reduction in memory with negligible loss in accuracy compared to the unpruned networks. This significantly reduces the computational resources required to train 3d CNNs. The pruned network obtained by our algorithm can also be easily scaled up and transferred to another dataset for training.

关键词： 3d CNNs 3d semantic segmentation neuron pruning resource efficient video classification

来源：评论

学校读者我要写书评

暂无评论

Towards Large-scale 3d Representation Learning with Multi-dataset Point Prompt Training

Towards Large-scale 3D Representation Learning with Multi-da...

引用

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Wu, Xiaoyang Tian, Zhuotao Wen, Xin Peng, Bohao Liu, Xihui Yu, Kaicheng Zhao, Hengshuang Univ Hong Kong Hong Kong Peoples R China Chinese Univ Hong Kong Hong Kong Peoples R China Westlake Univ Hangzhou Peoples R China Alibaba Grp Hangzhou Peoples R China

ISBN: (纸本)9798350353006

The rapid advancement of deep learning models is often attributed to their ability to leverage massive training data. In contrast, such privilege has not yet fully benefited 3d deep learning, mainly due to the limited availability of large-scale 3d datasets. Merging multiple available data sources and letting them collaboratively train a single model is a potential solution. However, due to the large domain gap between 3d point cloud datasets, such mixed supervision could adversely affect the model's performance and lead to degenerated performance (i.e., negative transfer) compared to single-dataset training. In view of this challenge, we introduce Point Prompt Training (PPT), a novel framework for multi-dataset synergistic learning in the context of 3d representation learning that supports multiple pre-training paradigms. Based on this framework, we propose Prompt-driven Normalization, which adapts the model to different datasets with domain-specific prompts and Language-guided Categorical Alignment that decently unifies the multiple-dataset label spaces by leveraging the relationship between label text. Extensive experiments verify that PPT can overcome the negative transfer associated with synergistic learning and produce generalizable representations. Notably, it achieves state-of-the-art performance on each dataset using a single weight-shared model with supervised multi-dataset training. Moreover, when served as a pre-training framework, it outperforms other pre-training approaches regarding representation quality and attains remarkable state-of-the-art performance across over ten diverse downstream tasks spanning both indoor and outdoor 3d scenarios.

关键词： 3d Pre-training 3d Representation Learning 3d semantic segmentation

来源：评论

学校读者我要写书评

暂无评论

Mix3d: Out-of-Context data Augmentation for 3d Scenes 9

Mix3D: Out-of-Context Data Augmentation for 3D Scenes

引用

9th International Conference on 3d Vision (3dV)

作者： Nekrasov, Alexey Schult, Jonas Litany, Or Leibe, Bastian Engelmann, Francis Rhein Westfal TH Aachen Aachen Germany NVIDIA Sunnyvale CA USA ETH AI Ctr Zurich Switzerland

ISBN: (纸本)9781665426886

We present Mix3d, a data augmentation technique for segmenting large-scale 3d scenes. Since scene context helps reasoning about object semantics, current works focus on models with large capacity and receptive fields that can fully capture the global context of an input 3d scene. However, strong contextual priors can have detrimental implications like mistaking a pedestrian crossing the street for a car. In this work, we focus on the importance of balancing global scene context and local geometry, with the goal of generalizing beyond the contextual priors in the training set. In particular, we propose a "mixing" technique which creates new training samples by combining two augmented scenes. By doing so, object instances are implicitly placed into novel out-of-context environments, therefore making it harder for models to rely on scene context alone, and instead infer semantics from local structure as well. We perform detailed analysis to understand the importance of global context, local structures and the effect of mixing scenes. In experiments, we show that models trained with Mix3d profit from a significant performance boost on indoor (ScanNet, S3dIS) and outdoor datasets (semanticKITTI). Mix3d can be trivially used with any existing method, e.g., trained with Mix3d, MinkowskiNet outperforms all prior state-of-the-art methods by a significant margin on the ScanNet test benchmark (78.1% mIoU). Code is available at: https://***/mix3d/

关键词： 3d data augmentation 3d scene understanding 3d semantic segmentation

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：