检索结果-内蒙古大学图书馆

1.2 Watt Classification of 3d Voxel Based Point-clouds using a CNN on a neural Compute Stick

NEUROCOMPUTING 2020年 393卷 165-174页

作者： Xu, Xiaofan Caulfield, Sam Amaro, Joao Falcao, Gabriel Moloney, david Intel Ireland Ltd Movidius Div Leixlip Kildare Ireland Univ Coimbra Inst Telecomunicacoes Dept Elect & Comp Engn Coimbra Portugal

With the recent surge in popularity of convolutional neural networks (CNNs), motivated by their significant performance in many classification and related tasks, a new challenge now needs to be addressed: how to accommodate CNNs in mobile devices, such as drones, smartphones, and similar low-power devices? In order to tackle this challenge we exploit the Vision Processing Unit (VPU) that combines dedicated CNN hardware blocks and very low power requirement. The lack of readily available training data and memory requirements are two of the factors hindering the training and accuracy performance of 3d CNNs. In this paper, we propose a method for generating synthetic 3d point-clouds from realistic CAd scene models to enrich the training process for volumetric CNNs. Furthermore, an efficient 3d volumetric object representation Volumetric Accelerator format (VOLA) is employed. VOLA is a sexaquaternary (power-of-four subdivision) tree-based representation which allows for significant memory saving for volumetric data. Multiple CNN models were trained and pruning techniques for the weights were applied to the trained 3d Volumetric network in order to remove almost 70% of the parameters and outperform the existing state-of-the-art networks. The top performing and efficient model was ported to the Movidius (TM) neural Compute Stick (NCS). After deployment on the NCS, it takes 11 ms (similar to 90 frames per second) to perform inference on each input volume, with a reported power requirement of 1.2W, which leads to 75.75 inferences per second per Watt. (C) 2019 Elsevier B.V. All rights reserved.

关键词： 3d convolutional neural network 3d object classification Embedded systems Low-power

来源：评论

学校读者我要写书评

暂无评论

Multi-Level Temporal dilated dense Prediction for Action Recognition

引用

IEEE TRANSACTIONS ON MULTIMEdIA 2022年 24卷 2553-2566页

作者： Wang, Jinpeng Lin, Yiqi Zhang, Manlin Gao, Yuan Ma, Andy J. Sun Yat Sen Univ Sch Elect & Informat Technol Guangzhou 510006 Peoples R China Sun Yat Sen Univ Sch Comp Sci & Engn Guangzhou 510006 Peoples R China

3d convolutional neural networks have achieved great success for action recognition. However, large variations of temporal dynamics have not been properly processed and low-level features have not been fully exploited in most existing works. To solve these two problems, we present a general and flexible framework, namely multi-level temporal dilated dense prediction network, which can incorporate with most of existing methods as backbone to improve the temporal modeling capacity. In the proposed method, a novel temporal dilated dense prediction block is designed to fully utilize temporal features with various temporal dilated rates for dense prediction while maintaining relatively low computational cost. To fuse information from low to high levels, our method combines the predictions from multiple such blocks inserted at different stages of the backbone network. In-depth analysis is given to show that short- to long-term temporal dependencies can be captured and multi-level spatio-temporal features are effectively fused for video action recognition by the proposed method. Experimental results demonstrate that our method achieves impressive performance improvement on four publicly available action recognition benchmarks including Charades, Kinetics, Something-Something-V1 and HMdB51.

关键词： Feature extraction Three-dimensional displays Convolution Image recognition Task analysis Solid modeling Predictive models Action Recognition Temporal dilated dense Prediction Multi-level Fusion 3d convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

A novel online action detection framework from untrimmed video streams

引用

PATTERN RECOGNITION 2020年 106卷 107396-107396页

作者： Yoon, da-Hye Cho, Nam-Gyu Lee, Seong-Whan Korea Univ Dept Comp & Radio Commun Engn Seoul 02841 South Korea Korea Univ Dept Brain & Cognit Engn Seoul 02841 South Korea Korea Univ Dept Artificial Intelligence Seoul 02841 South Korea

Online temporal action localization from an untrimmed video stream is a challenging problem in computer vision. It is challenging because of i) in an untrimmed video stream, more than one action instance may appear, including background scenes, and ii) in online settings, only past and current information is available. Therefore, temporal priors, such as the average action duration of training data, which have been exploited by previous action detection methods, are not suitable for this task because of the high intra-class variation in human actions. We propose a novel online action detection framework that considers actions as a set of temporally ordered subclasses and leverages a future frame generation network to cope with the limited information issue associated with the problem outlined above. Additionally, we augment our data by varying the lengths of videos to allow the proposed method to learn about the high intra-class variation in human actions. We evaluate our method using two benchmark datasets, THUMOS'14 and ActivityNet, for an online temporal action localization scenario and demonstrate that the performance is comparable to state-of-the-art methods that have been proposed for offline settings. (C) 2020 Elsevier Ltd. All rights reserved.

关键词： Online action detection Untrimmed video stream Future frame generation 3d convolutional neural network Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Multiple damage detection of maglev rail joints using time-frequency spectrogram and convolutional neural network

引用

SMART STRUCTURES ANd SYSTEMS 2022年第4期29卷 625-640页

作者： Wang, Su-Mei Jiang, Gao-Feng Ni, Yi-Qing Lu, Yang Lin, Guo-Bin Pan, Hong-Liang Xu, Jun-Qi Hao, Shuo Hong Kong Polytech Univ Dept Civil & Environm Engn Hung Hom Kowloon Hong Kong Peoples R China Natl Rail Transit Electrificat & Automat Engn Tec Hong Kong Branch Hung Hom Kowloon Hong Kong Peoples R China Tongji Univ Shanghai Key Lab Rail Infrastruct Durabil & Syst Shanghai Peoples R China Tongji Univ Maglev Transportat Engn R&D Ctr Shanghai Peoples R China

Maglev rail joints are vital components serving as connections between the adjacent F-type rail sections in maglev guideway. damage to maglev rail joints such as bolt looseness may result in rough suspension gap fluctuation, failure of suspension control, and even sudden clash between the electromagnets and F-type rail. The condition monitoring of maglev rail joints is therefore highly desirable to maintain safe operation of maglev. In this connection, an online damage detection approach based on three-dimensional (3d) convolutional neural network (CNN) and time-frequency characterization is developed for simultaneous detection of multiple damage of maglev rail joints in this paper. The training and testing data used for condition evaluation of maglev rail joints consist of two months of acceleration recordings, which were acquired in-situ from different rail joints by an integrated online monitoring system during a maglev train running on a test line. Short-time Fourier transform (STFT) method is applied to transform the raw monitoring data into time-frequency spectrograms (TFS). Three CNN architectures, i.e., small-sized CNN (S-CNN), middle-sized CNN (M-CNN), and large-sized CNN (L-CNN), are configured for trial calculation and the M-CNN model with excellent prediction accuracy and high computational efficiency is finally optioned for multiple damage detection of maglev rail joints. Results show that the rail joints in three different conditions (bolt-looseness caused rail step, misalignment-caused lateral dislocation, and normal condition) are successfully identified by the proposed approach, even when using data collected from rail joints from which no data were used in the CNN training. The capability of the proposed method is further examined by using the data collected after the loosed bolts have been replaced. In addition, by comparison with the results of CNN using frequency spectrum and traditional neural network using TFS, the proposed TFSCNN framework is

关键词： 3d convolutional neural network damage detection maglev rail joints time-frequency spectrogram

来源：评论

学校读者我要写书评

暂无评论

3d-CNN-based feature extraction of ground-based cloud images for direct normal irradiance prediction

引用

SOLAR ENERGY 2019年 181卷 510-518页

作者： Zhao, Xin Wei, Haikun Wang, Hai Zhu, Tingting Zhang, Kanjian Southeast Univ Sch Automat Minist Educ Key Lab Measurement & ControlCSE Nanjing 210096 Jiangsu Peoples R China St Marys Univ Sobey Sch Business Halifax NS Canada

Cloud cover and cloud motion have a large impact on solar irradiance. One of the effective ways for direct normal irradiance (dNI) prediction is to use cloud features, which has been extensively studied. A convolutional neural network (CNN) has the advantage of automatic features extraction by using strong computing capabilities. In this paper, a novel 3d-CNN method is proposed by processing multiple consecutive ground-based cloud (GBC) images in order to extract cloud features including texture and temporal information. The resulting features and the dNI data are then used to establish a dNI forecasting model. The experiments are carried out to evaluate the performance of the proposed forecasting method by using the data from January 1, 2013 to december 31, 2014. The experimental results show that the proposed method coupled with the multilayer perceptron (MLP) model achieves forecast skill of 17.06% for 10-minute ahead dNI prediction.

关键词： direct normal irradiance 3d convolutional neural network Feature extraction Ground-based cloud image

来源：评论

学校读者我要写书评

暂无评论

Three-stream spatio-temporal attention network for first-person action and interaction recognition

引用

JOURNAL OF AMBIENT INTELLIGENCE ANd HUMANIZEd COMPUTING 2022年第2期13卷 1137-1152页

作者： Imran, Javed Raman, Balasubramanian Univ Petr & Energy Studies Dept Informat Sch Comp Sci Dehra Dun Uttarakhand India Indian Inst Technol Roorkee Dept Comp Sci & Engn Roorkee Uttar Pradesh India

The problem of action and interaction recognition of human activities from the perspective of first-person view-point is an interesting area of research in the field of human action recognition (HAR). This paper presents a data-driven spatio-temporal network to combine different modalities computed from first-person videos using a temporal attention mechanism. First, our proposed approach uses three-stream inflated 3d ConvNet (I3d) to extract low-level features from RGB frame difference (Fd), optical flow (OF) and magnitude-orientation (MO) streams. An I3d network has the advantage to directly learn spatio-temporal features over short video snippets (like 16 frames). Second, the extracted features are fused together and fed to a Bidirectional long short-term memory (BiLSTM) network to model high-level temporal feature sequences. Third, we propose to incorporate attention mechanism with our BiLSTM network to automatically select the most relevant temporal snippets in the given video sequence. Finally, we conducted extensive experiments and achieve state-of-the-art results on JPL (98.5%), NUS (84.1%), UTK (91.5%) and dogCentric (83.3%) datasets. These results show that features extracted from three-stream network are complementary to each other, and attention mechanism further improves the results by a large margin than previous attempts based on handcrafted and deep features.

关键词： First-person action recognition 3d convolutional neural network Recurrent neural network Feature fusion Soft attention

来源：评论

学校读者我要写书评

暂无评论

Efficient multi-scale 3d CNN with fully connected CRF for accurate brain lesion segmentation

引用

MEdICAL IMAGE ANALYSIS 2017年 36卷 61-78页

作者： Kamnitsas, Konstantinos Ledig, Christian Newcombe, Virginia F. J. Sirnpson, Joanna P. Kane, Andrew d. Menon, david K. Rueckert, daniel Glocker, Ben Imperial Coll London Biomed Image Anal Grp London England Univ Cambridge Univ Div Anaesthesia Dept Med Cambridge CB2 1TN England Univ Cambridge Wolfson Brain Imaging Ctr Cambridge CB2 1TN England

We propose a dual pathway, 11-layers deep, three-dimensional convolutional neural network for the challenging task of brain lesion segmentation. The devised architecture is the result of an in-depth analysis of the limitations of current networks proposed for similar applications. To overcome the computational burden of processing 3d medical scans, we have devised an efficient and effective dense training scheme which joins the processing of adjacent image patches into one pass through the network while automatically adapting to the inherent class imbalance present in the data. Further, we analyze the development of deeper, thus more discriminative 3d CNNs. In order to incorporate both local and larger contextual information, we employ a dual pathway architecture that processes the input images at multiple scales simultaneously. For post-processing of the network's soft segmentation, we use a 3d fully connected Conditional Random Field which effectively removes false positives. Our pipeline is extensively evaluated on three challenging tasks of lesion segmentation in multi-channel MRI patient data with traumatic brain injuries, brain tumours, and ischemic stroke. We improve on the state-of-the-art for all three applications, with top ranking performance on the public benchmarks BRATS 2015 and ISLES 2015. Our method is computationally efficient, which allows its adoption in a variety of research and clinical settings. The source code of our implementation is made publicly available. (C) 2016 The Authors. Published by Elsevier B.V.

关键词： 3d convolutional neural network Fully connected CRF Segmentation Brain lesions deep learning

来源：评论

学校读者我要写书评

暂无评论

deep Action Parsing in Videos With Large-Scale Synthesized data

引用

IEEE TRANSACTIONS ON IMAGE PROCESSING 2018年第6期27卷 2869-2882页

作者： Liu, Li Zhou, Yi Shao, Ling Incept Inst Artificial Intelligence Abu Dhabi U Arab Emirates

Action parsing in videos with complex scenes is an interesting but challenging task in computer vision. In this paper, we propose a generic 3d convolutional neural network in a multi-task learning manner for effective deep Action Parsing (dAP3d-Net) in videos. Particularly, in the training phase, action localization, classification, and attributes learning can be jointly optimized on our appearance-motion data via dAP3d-Net. For an upcoming test video, we can describe each individual action in the video simultaneously as: Where the action occurs, What the action is, and How the action is performed. To well demonstrate the effectiveness of the proposed dAP3d-Net, we also contribute a new Numerous-category Aligned Synthetic Action data set, i.e., NASA, which consists of 200 000 action clips of over 300 categories and with 33 pre-defined action attributes in two hierarchical levels (i.e., low-level attributes of basic body part movements and high-level attributes related to action motion). We learn dAP3d-Net using the NASA data set and then evaluate it on our collected Human Action Understanding data set and the public THUMOS data set. Experimental results show that our approach can accurately localize, categorize, and describe multiple actions in realistic videos.

关键词： Action parsing 3d convolutional neural network action localization and classification attributes learning NASA and HAU

来源：评论

学校读者我要写书评

暂无评论

Wireless capsule endoscopy multiclass classification using three-dimensional deep convolutional neural network model

引用

BIOMEdICAL ENGINEERING ONLINE 2023年第1期22卷 1-23页

作者： Bordbar, Mehrdokht Helfroush, Mohammad Sadegh danyali, Habibollah Ejtehadi, Fardad Shiraz Univ Technol Dept Elect Engn Shiraz Iran Shiraz Univ Med Sci Gastroenterohepatol Res Ctr Sch Med Dept Internal Med Shiraz Iran

BackgroundWireless capsule endoscopy (WCE) is a patient-friendly and non-invasive technology that scans the whole of the gastrointestinal tract, including difficult-to-access regions like the small bowel. Major drawback of this technology is that the visual inspection of a large number of video frames produced during each examination makes the physician diagnosis process tedious and prone to error. Several computer-aided diagnosis (CAd) systems, such as deep network models, have been developed for the automatic recognition of abnormalities in WCE frames. Nevertheless, most of these studies have only focused on spatial information within individual WCE frames, missing the crucial temporal data within consecutive *** this article, an automatic multiclass classification system based on a three-dimensional deep convolutional neural network (3d-CNN) is proposed, which utilizes the spatiotemporal information to facilitate the WCE diagnosis process. The 3d-CNN model fed with a series of sequential WCE frames in contrast to the two-dimensional (2d) model, which exploits frames as independent ones. Moreover, the proposed 3d deep model is compared with some pre-trained networks. The proposed models are trained and evaluated with 29 subject WCE videos (14,691 frames before augmentation). The performance advantages of 3d-CNN over 2d-CNN and pre-trained networks are verified in terms of sensitivity, specificity, and ***3d-CNN outperforms the 2d technique in all evaluation metrics (sensitivity: 98.92 vs. 98.05, specificity: 99.50 vs. 86.94, accuracy: 99.20 vs. 92.60). In conclusion, a novel 3d-CNN model for lesion detection in WCE frames is proposed in this *** results indicate the performance of 3d-CNN over 2d-CNN and some well-known pre-trained classifier networks. The proposed 3d-CNN model uses the rich temporal information in adjacent frames as well as spatial data to develop an accurate and efficient model.

关键词： Wireless capsule endoscopy Image classification deep learning 3d convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

An optimal model for identification and classification of corn leaf disease using hybrid 3d-CNN and LSTM

引用

BIOMEdICAL SIGNAL PROCESSING ANd CONTROL 2024年 92卷

作者： Ashwini, C. Sellam, V. SRM Inst Sci & Technol Dept Comp Sci & Engn Ramapuram Tamil Nadu India

More than 30 % of corn is produced by the United States, which is the world's leading corn producer. The deficiency in the corn leaf may influence crop yield manufacturing. Moreover, those methods only serve yield assessment and do not compute the prediction uncertainties in the crop. To overcome such limitations, the hybrid model called Whale Optimization Algorithm with Joint Search Mechanisms (JSWOA) optimized hybrid 3dCNNRNN model is proposed for the corn leaf disease prediction. 3dCNN includes Four MaxPooling3d layers and six 3d convolutional layers to enhance the identification of 3-dimensional images, and it determines the temporally local features and spatial features. JSWOA optimized hyperparameters of the hybrid 3dCNN-RNN models. For performing different experiments based on machine learning, Maize_in_field, and KaraAgro AI maize are the two datasets employed. The performance rate of MCC, recall, precision, accuracy, specificity, F1-score, and AUC are assessed to ensure the efficacy of the proposed hybrid model. Also, the existing CNN, optimized denseNet, and AlexNet models are analyzed for performance comparison. The simulation outcome reveals that the proposed hybrid model reaches above 90 % performance in predicting various maize leaf classes on both datasets. Therefore, the state-of-the-art method is outperformed by the proposed hybrid model.

关键词： Corn leaf disease prediction Whale optimization algorithm with Joint search Mechanisms 3d convolutional neural network Long Short -Term Memory

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：