检索结果-内蒙古大学图书馆

arXiv 2019年

作者： Hu, Wenqing Zhu, Zhanxing Xiong, Haoyi Huan, Jun Department of Mathematics and Statistics Missouri University of Science and Technology University of Missouri Rolla Peking University Beijing Institute of Big Data Research Beijing China Big Data Lab Baidu Inc. National Engineering Laboratory of Deep Learning Application and Technology

We interpret the variational inference of the Stochastic Gradient Descent (SGD) as minimizing a new potential function named the quasi-potential. We analytically construct the quasi-potential function in the case when the loss function is convex and admits only one global minimum point. We show in this case that the quasi-potential function is related to the noise covariance structure of SGD via a partial differential equation of Hamilton-Jacobi type. This relation helps us to show that anisotropic noise leads to faster escape than isotropic noise. We then consider the dynamics of SGD in the case when the loss function is non-convex and admits several different local minima. In this case, we demonstrate an example that shows how the noise covariance structure plays a role in "implicit regularization", a phenomenon in which SGD favors some particular local minimum points. This is done through the relation between the noise covariance structure and the quasi-potential function. Our analysis is based on Large Deviations Theory (LDT), and they are validated by numerical experiments. Copyright © 2019, The Authors. All rights reserved.

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

Joint 3D Instance Segmentation and Object Detection for Autonomous Driving

Joint 3D Instance Segmentation and Object Detection for Auto...

引用

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Dingfu Zhou Jin Fang Xibin Song Liu Liu Junbo Yin Yuchao Dai Hongdong Li Ruigang Yang Baidu Research National Engineering Laboratory of Deep Learning Technology and Application Beijing China Australian National University Canberra Australia Australian Centre for Robotic Vision Australia Beijing Institute of Technology Beijing China Northwestern Polytechnical University Xi'an China University of Kentucky Kentucky USA

ISBN: (数字)9781728171685

ISBN: (纸本)9781728171692

Currently, in Autonomous Driving (AD), most of the 3D object detection frameworks (either anchor- or anchor-free-based) consider the detection as a Bounding Box (BBox) regression problem. However, this compact representation is not sufficient to explore all the information of the objects. To tackle this problem, we propose a simple but practical detection framework to jointly predict the 3D BBox and instance segmentation. For instance segmentation, we propose a Spatial Embeddings (SEs) strategy to assemble all foreground points into their corresponding object centers. Base on the SE results, the object proposals can be generated based on a simple clustering strategy. For each cluster, only one proposal is generated. Therefore, the Non-Maximum Suppression (NMS) process is no longer needed here. Finally, with our proposed instance-aware ROI pooling, the BBox is refined by a second-stage network. Experimental results on the public KITTI dataset show that the proposed SEs can significantly improve the instance segmentation results compared with other feature embedding-based method. Meanwhile, it also outperforms most of the 3D object detectors on the KITTI testing benchmark.

关键词： Three-dimensional displays Object detection Two dimensional displays Proposals Feature extraction Shape Semantics

来源：评论

学校读者我要写书评

暂无评论

A Unified Object Motion and Affinity Model for Online Multi-Object Tracking

A Unified Object Motion and Affinity Model for Online Multi-...

引用

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Junbo Yin Wenguan Wang Qinghao Meng Ruigang Yang Jianbing Shen Beijing Lab of Intelligent Information Technology School of Computer Science Beijing Institute of Technology China ETH Zurich Switzerland Baidu Research National Engineering Laboratory of Deep Learning Technology and Application China University of Kentucky Kentucky USA Inception Institute of Artificial Intelligence UAE

ISBN: (数字)9781728171685

ISBN: (纸本)9781728171692

Current popular online multi-object tracking (MOT) solutions apply single object trackers (SOTs) to capture object motions, while often requiring an extra affinity network to associate objects, especially for the occluded ones. This brings extra computational overhead due to repetitive feature extraction for SOT and affinity computation. Meanwhile, the model size of the sophisticated affinity network is usually non-trivial. In this paper, we propose a novel MOT framework that unifies object motion and affinity model into a single network, named UMA, in order to learn a compact feature that is discriminative for both object motion and affinity measure. In particular, UMA integrates single object tracking and metric learning into a unified triplet network by means of multi-task learning. Such design brings advantages of improved computation efficiency, low memory requirement and simplified training procedure. In addition, we equip our model with a task-specific attention module, which is used to boost task-aware feature learning. The proposed UMA can be easily trained end-to-end, and is elegant - requiring only one training stage. Experimental results show that it achieves promising performance on several MOT Challenge benchmarks.

关键词： Computational modeling Task analysis Target tracking Feature extraction Training

来源：评论

学校读者我要写书评

暂无评论

View extrapolation of human body from a single image

arXiv

引用

arXiv 2018年

作者： Zhu, Hao Su, Hao Wang, Peng Cao, Xun Yang, Ruigang Nanjing University Nanjing China University of Kentucky LexingtonKY United States University of California San DiegoCA United States Baidu Inc. Beijing China National Engineering Laboratory of Deep Learning and Technology and Application China

We study how to synthesize novel views of human body from a single image. Though recent deep learning based methods work well for rigid objects, they often fail on objects with large articulation, like human bodies. The core step of existing methods is to fit a map from the observable views to novel views by CNNs;however, the rich articulation modes of human body make it rather challenging for CNNs to memorize and interpolate the data well. To address the problem, we propose a novel deep learning based pipeline that explicitly estimates and leverages the geometry of the underlying human body. Our new pipeline is a composition of a shape estimation network and an image generation network, and at the interface a perspective transformation is applied to generate a forward flow for pixel value transportation. Our design is able to factor out the space of data variation and makes learning at each step much easier. Empirically, we show that the performance for pose-varying objects can be improved dramatically. Our method can also be applied on real data captured by 3D sensors, and the flow generated by our methods can be used for generating high quality results in higher resolution. Copyright © 2018, The Authors. All rights reserved.

关键词： Pipelines

来源：评论

学校读者我要写书评

暂无评论

LiDAR-based online 3D video object detection with graph-based message passing and spatiotemporal transformer attention

arXiv

引用

arXiv 2020年

作者： Yin, Junbo Shen, Jianbing Guan, Chenye Zhou, Dingfu Yang, Ruigang Beijing Lab of Intelligent Information Technology School of Computer Science Beijing Institute of Technology China Baidu Research National Engineering Laboratory of Deep Learning Technology and Application China Inception Institute of Artificial Intelligence United Arab Emirates University of Kentucky Kentucky United States

Existing LiDAR-based 3D object detectors usually focus on the single-frame detection, while ignoring the spatiotemporal information in consecutive point cloud frames. In this paper, we propose an end-to-end online 3D video object detector that operates on point cloud sequences. The proposed model comprises a spatial feature encoding component and a spatiotemporal feature aggregation component. In the former component, a novel Pillar Message Passing Network (PMPNet) is proposed to encode each discrete point cloud frame. It adaptively collects information for a pillar node from its neighbors by iterative message passing, which effectively enlarges the receptive field of the pillar feature. In the latter component, we propose an Attentive Spatiotemporal Transformer GRU (AST-GRU) to aggregate the spatiotemporal information, which enhances the conventional ConvGRU with an attentive memory gating mechanism. AST-GRU contains a Spatial Transformer Attention (STA) module and a Temporal Transformer Attention (TTA) module, which can emphasize the foreground objects and align the dynamic objects, respectively. Experimental results demonstrate that the proposed 3D video object detector achieves state-of-the-art performance on the large-scale nuScenes benchmark. Copyright © 2020, The Authors. All rights reserved.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Rethinking Table Recognition using Graph Neural Networks

Rethinking Table Recognition using Graph Neural Networks

引用

International Conference on Document Analysis and Recognition

作者： Shah Rukh Qasim Hassan Mahmood Faisal Shafait School of Electrical Engineering and Computer Science (SEECS) National University of Sciences and Technology (NUST) Islamabad Pakistan Deep Learning Laboratory National Center of Artificial Intelligence (NCAI) Islamabad Pakistan

Document structure analysis, such as zone segmentation and table recognition, is a complex problem in document processing and is an active area of research. The recent success of deep learning in solving various computer vision and machine learning problems has not been reflected in document structure analysis since conventional neural networks are not well suited to the input structure of the problem. In this paper, we propose an architecture based on graph networks as a better alternative to standard neural networks for table recognition. We argue that graph networks are a more natural choice for these problems, and explore two gradient-based graph neural networks. Our proposed architecture combines the benefits of convolutional neural networks for visual feature extraction and graph networks for dealing with the problem structure. We empirically demonstrate that our method outperforms the baseline by a significant margin. In addition, we identify the lack of large scale datasets as a major hindrance for deep learning research for structure analysis and present a new large scale synthetic dataset for the problem of table recognition. Finally, we open-source our implementation of dataset generation and the training framework of our graph networks to promote reproducible research in this direction.

关键词： Feature extraction Computer architecture Convolutional neural networks Text analysis deep learning Training

来源：评论

学校读者我要写书评

暂无评论

Table Structure Extraction with Bi-Directional Gated Recurrent Unit Networks

Table Structure Extraction with Bi-Directional Gated Recurre...

引用

International Conference on Document Analysis and Recognition

作者： Saqib Ali Khan Syed Muhammad Daniyal Khalid Muhammad Ali Shahzad Faisal Shafait School of Electrical Engineering and Computer Science (SEECS) National University of Sciences and Technology (NUST) Islamabad Pakistan Deep Learning Laboratory National Center of Artificial Intelligence (NCAI) Islamabad Pakistan

Tables present summarized and structured information to the reader, which makes table's structure extraction an important part of document understanding applications. However, table structure identification is a hard problem not only because of the large variation in the table layouts and styles, but also owing to the variations in the page layouts and the noise contamination levels. A lot of research has been done to identify table structure, most of which is based on applying heuristics with the aid of optical character recognition (OCR) to hand pick layout features of the tables. These methods fail to generalize well because of the variations in the table layouts and the errors generated by OCR. In this paper, we have proposed a robust deep learning based approach to extract rows and columns from a detected table in document images with a high precision. In the proposed solution, the table images are first pre-processed and then fed to a bi-directional Recurrent Neural Network with Gated Recurrent Units (GRU) followed by a fully-connected layer with softmax activation. The network scans the images from top-to-bottom as well as left-to-right and classifies each input as either a row-separator or a column-separator. We have benchmarked our system on publicly available UNLV as well as ICDAR 2013 datasets on which it outperformed the state-of-theart table structure extraction systems by a significant margin.

关键词： Layout Bidirectional control Image segmentation deep learning Recurrent neural networks Logic gates Optical character recognition software

来源：评论

学校读者我要写书评

暂无评论

Incremental learning of Object Detector with Limited Training Data

Incremental Learning of Object Detector with Limited Trainin...

引用

Proceedings of the Digital Image Computing: Technqiues and applications (DICTA)

作者： Muhammad Abdullah Hafeez Adnan Ul-Hasan Faisal Shafait School of Electrical Engineering and Computer Science (SEECS) National University of Sciences and Technology (NUST) Islamabad Pakistan Deep Learning Laboratory National Center of Artificial Intelligence (NCAI) Islamabad Pakistan

State of the art deep learning models, despite being at par to the human level in some of the challenging tasks, still suffer badly when they are put in the condition where they have to learn with time. This open challenge problem of making deep learning model learn with time is referred in the literature as Lifelong learning, Incremental learning or Continual learning. In each increment, new classes/tasks are introduced to the existing model and trained on them while maintaining the accuracy of the previously learned classes/tasks. But accuracy of the deep learning model on the previously learned classes/tasks decreases with each increment. The main reason behind this accuracy drop is catastrophic forgetting, an inherent flaw in the deep learning models, where weights learned during the past increments, get disturbed while learning the new classes/tasks from new increment. Several approaches have been proposed to mitigate or avoid this catastrophic forgetting, such as the use of knowledge distillation, rehearsal over previous classes, or dedicated paths for different increments, etc. In this work, we have proposed a novel approach based on transfer learning methodology, which uses a combination of pre-trained shared and fixed network as a backbone, along with a dedicated network extension in incremental setting for the learning of new tasks incrementally. The results have shown that our approach has better performance in two ways. First, our model has significantly better overall incremental accuracy than that of the best in class model in different incremental configurations. Second, our approach achieves better results while maintaining properties of true incremental learning algorithm i.e. successful avoidance of the catastrophic forgetting issue and complete eradication of the need of saved exemplars or retraining phases, which are required by the current state of the art model to maintain performance.

关键词： deep learning Training Knowledge engineering Philosophical considerations Digital images Transfer learning Training data

来源：评论

学校读者我要写书评

暂无评论

Enhancing Multimodal Information Extraction from Visually Rich Documents with 2D Positional Embeddings

Enhancing Multimodal Information Extraction from Visually Ri...

引用

Proceedings of the Digital Image Computing: Technqiues and applications (DICTA)

作者： Aresha Arshad Momina Moetesum Adnan Ul Hasan Faisal Shafait School of Electrical Engineering and Computer Science (SEECS) National University of Sciences and Technology (NUST) Islamabad Pakistan Deep Learning Laboratory National Center of Artificial Intelligence (NCAI) Islamabad Pakistan

ISBN: (数字)9798350379037

ISBN: (纸本)9798350379044

Visually rich document understanding involves the interpretation of documents with varied formats and complex layouts, including multi-line entities, presenting a significant challenge. This study addresses these challenges by introducing a document comprehension model based on the LayoutLMv3 architecture, incorporating two-dimensional (2D) positional embeddings to capture both row and column information. Additionally, a multi-stage Transformer network is employed for hierarchical processing of document features. The proposed model is evaluated through extensive experiments on FUNSD dataset, achieving improved performance in both text-centric and image-centric document understanding tasks. Results demonstrate enhanced spatial comprehension and computational efficiency as compared to the state-of-the-art, establishing our approach as a significant contribution to the field of visually rich document understanding.

关键词： Computational modeling Two-dimensional displays Digital images Layout Computer architecture Transformers Information retrieval Encoding Computational efficiency Usability

来源：评论

学校读者我要写书评

暂无评论

Towards Making deep Transfer learning Never Hurt

Towards Making Deep Transfer Learning Never Hurt

引用

IEEE International Conference on Data Mining (ICDM)

作者： Ruosi Wan Haoyi Xiong Xingjian Li Zhanxing Zhu Jun Huan Big Data Laboratory Baidu Inc. Beijing China School of Mathematical Sciences Peking University Beijing China National Engineering Laboratory for Deep Learning Technology and Applications Beijing China

Transfer learning have been frequently used to improve deep neural network training through incorporating weights of pre-trained networks as the starting-point of optimization for regularization. While deep transfer learning can usually boost the performance with better accuracy and faster convergence, transferring weights from inappropriate networks hurts training procedure and may lead to even lower accuracy. In this paper, we consider deep transfer learning as minimizing a linear combination of empirical loss and regularizer based on pre-trained weights, where the regularizer would restrict the training procedure from lowering the empirical loss, with conflicted descent directions (e.g., derivatives). Following the view, we propose a novel strategy making regularization-based deep Transfer learning Never Hurt (DTNH) that, for each iteration of training procedure, computes the derivatives of the two terms separately, then re-estimates a new descent direction that does not hurt the empirical loss minimization while preserving the regularization affects from the pre-trained weights. Extensive experiments have been done using common transfer learning regularizers, such as L2-SP and knowledge distillation, on top of a wide range of deep transfer learning benchmarks including Caltech, MIT indoor 67, CIFAR-10 and ImageNet. The empirical results show that the proposed descent direction estimation strategy DTNH can always improve the performance of deep transfer learning tasks based on all above regularizers, even when transferring pre-trained weights from inappropriate networks. All in all, DTNH strategy can improve state-of-the-art regularizers in all cases with 0.1%-7% higher accuracy in all experiments.

关键词：

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：