检索结果-内蒙古大学图书馆

arXiv 2019年

作者： Chaturvedi, Akshay Abijith, K.P. Garain, Utpal Computer Vision and Pattern Recognition Unit Indian Statistical Institute India

Neural machine translation (NMT) systems have been shown to give undesirable translation when a small change is made in the source sentence. In this paper, we study the behaviour of NMT systems when multiple changes are made to the source sentence. In particular, we ask the following question "Is it possible for an NMT system to predict same translation even when multiple words in the source sentence have been replaced?". To this end, we propose a soft-attention based technique to make the aforementioned word replacements. The experiments are conducted on two language pairs: English- German (en-de) and English-French (en-fr) and two state-of-theart NMT systems: BLSTM-based encoder-decoder with attention and Transformer. The proposed soft-attention based technique achieves high success rate and outperforms existing methods like HotFlip by a significant margin for all the conducted experiments. The results demonstrate that state-of-the-art NMT systems are unable to capture the semantics of the source language. The proposed soft-attention based technique is an invariance-based adversarial attack on NMT systems. To better evaluate such attacks, we propose an alternate metric and argue its benefits in comparison with success rate. Copyright © 2019, The Authors. All rights reserved.

关键词： Neural machine translation

来源：评论

学校读者我要写书评

暂无评论

A Holistic Approach for Recognition of Complete Urdu Ligatures Using Hidden Markov Models

A Holistic Approach for Recognition of Complete Urdu Ligatur...

引用

Frontiers of Information Technology (FIT)

作者： Israr Uddin Imran Siddiqi Shehzad Khalid Center of Computer Vision and Pattern Recognition Bahria University Islamabad Pakistan

Optical Character Recognition (OCR) is one of the continuously explored problems. Presently, commercial character recognizers are available reporting near to 100% recognition rates on text in a number of scripts. Despite these advancements, OCR systems however, have yet to mature for cursive scripts like Urdu. This study presents a holistic technique for recognition of Urdu text in Nastaliq font using "complete" ligatures as recognition units. The term "complete" refers to a partial word including its main body and secondary components (dots and diacritic marks). Discrete Wavelet Transform (DWT) is employed as feature extractor while a separate Hidden Markov Model (HMM) is trained for each ligature considered in our study. More than 2000 frequently used unique Urdu ligatures from the standard CLE (Center of Language Engineering) dataset are considered in our evaluations. The system reads a promising accuracy of 88.87% on more than 10,000 partial words.

关键词： Character recognition Feature extraction Hidden Markov models Text recognition Discrete wavelet transforms Training Optical character recognition software

来源：评论

学校读者我要写书评

暂无评论

Guiding Pseudo-labels with Uncertainty Estimation for Source-free Unsupervised Domain Adaptation

Guiding Pseudo-labels with Uncertainty Estimation for Source...

引用

Conference on computer vision and pattern Recognition (CVPR)

作者： Mattia Litrico Alessio Del Bue Pietro Morerio Pattern Analysis and Computer Vision (PAVIS) Istituto Italiano di Tecnologia

Standard Unsupervised Domain Adaptation (UDA) methods assume the availability of both source and target data during the adaptation. In this work, we investigate Source-free Unsupervised Domain Adaptation (SF-UDA), a specific case of UDA where a model is adapted to a target domain without access to source data. We propose a novel approach for the SF-UDA setting based on a loss reweighting strategy that brings robustness against the noise that inevitably affects the pseudo-labels. The classification loss is reweighted based on the reliability of the pseudo-labels that is measured by estimating their uncertainty. Guided by such reweighting strategy, the pseudo-labels are progressively refined by aggregating knowledge from neighbouring samples. Furthermore, a self-supervised contrastive framework is leveraged as a target space regulariser to enhance such knowledge aggregation. A novel negative pairs exclusion strategy is proposed to identify and exclude negative pairs made of samples sharing the same class, even in presence of some noise in the pseudo-labels. Our method outperforms previous methods on three major benchmarks by a large margin. We set the new SF-UDA state-of-the-art on VisDA-C and DomainNet with a performance gain of + 1.8% on both benchmarks and on PACS with + 12.3% in the single-source setting and +6.6% in multi-target adaptation. Additional analyses demonstrate that the proposed approach is robust to the noise, which results in significantly more accurate pseudo-labels compared to state-of-the-art approaches.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Offline handwritten Devanagari word recognition: Information fusion at feature and classifier levels

Offline handwritten Devanagari word recognition: Information...

引用

Asian Conference on pattern Recognition (ACPR)

作者： Bikash Shaw Ujjwal Bhattacharya Swapan Kumar Parui Computer Vision and Pattern Recognition Unit Indian Statistical Institute Kolkata

ISBN: (纸本)9781479961016

This article presents our recent study on fusion of information at feature and classifier output levels for improved performance of offline handwritten Devanagari word recognition. We consider here two state-of-the-art features, viz., Directional Distance Distribution (DDD) and Gradient-Structural-Concavity (GSC) features along with multi-class SVM classifiers. Here, we study various combinations of DDD features along with one or more features from the GSC feature set. We experiment by presenting different combined feature vectors as input to SVM classifiers. Also, the output vectors of different SVM classifiers fed with different feature vectors are combined by another SVM classifier. The combination of the outputs of two SVMs each being fed with a different feature vector provides superior performance to the performance of a single SVM classifier fed with the combined feature vector. Experimental results are obtained on a large handwritten Devanagari word sample image database of 100 Indian town names. The recognition results on its test samples show that SVM recognition output of DDD features combined with the SVM output of GSC features improves the final recognition accuracy significantly.

关键词： Handwriting recognition Support vector machines Feature extraction Shape Databases Automation

来源：评论

学校读者我要写书评

暂无评论

Robust stereo on multiple resolutions

Robust stereo on multiple resolutions

引用

13th International Conference on pattern Recognition, ICPR 1996

作者： Menard, Christian Leonardis, Aleš Department for Pattern Recognition and Image Processing Technical University Vienna Treitlstraße 3/1832 3A-1040 Vienna Austria University of Ljubljana Faculty of Computer and Information Science Computer Vision Laboratory Ljubljana Slovenia

ISBN: (纸本)081867282X

Stereo computation is one of the vision problems where the presence of outliers cannot be neglected. Most standard algorithms make unrealistic assumptions about noise distributions, which leads to erroneous results that cannot be corrected in subsequent postprocessing stages. In this paper we present a modification of the standard area-based correlation approach so that it can tolerate a significant number of outliers. The approach exhibits a robust behavior not only in the presence of mismatches but also in the case of depth discontinuities. The confidence measure of the correlation and the number of outliers provide two complementary sources of information which, when implemented in a multiresolution framework, result in a robust and efficient method. We present the results of this approach on a number of synthetic and real images. © 1996 IEEE.

关键词： Stereo image processing

来源：评论

学校读者我要写书评

暂无评论

Word-wise handwritten Persian and Roman script identification

Word-wise handwritten Persian and Roman script identificatio...

引用

International Conference on Frontiers in Handwriting Recognition

作者： Roy, Kaushik Alaei, Alireza Pal, Umapada Department of Computer Science West Bengal State University Kolkata-126 India Department of Studies in Computer Science University of Mysore Mysore 570 006 India Computer Vision and Pattern Recognition Unit Indian Statistical Institute Kolkata-108 India

ISBN: (纸本)9780769542218

Most of the countries use bi-script documents. This is because every country uses its own national language and English as second/foreign language. Therefore, bi-lingual document with one language being the English and other being the national language is very common. Postal documents are a very good example of such bi-lingual/script document. This paper deals with word-wise handwritten script identification from bi-script documents written in Persian and Roman. In the proposed scheme, simple but fast computable set of 12 features based on fractal dimension, position of small component, topology etc. are used and a set of classifiers are employed for script identification experiments. We tested our scheme on a dataset of 5000 handwritten Persian and English words and 99.20% of correct script identification is obtained. © 2010 IEEE.

关键词： Fractal dimension

来源：评论

学校读者我要写书评

暂无评论

Mimic and fool: a task agnostic adversarial attack

arXiv

引用

arXiv 2019年

作者： Chaturvedi, Akshay Garain, Utpal Computer Vision and Pattern Recognition Unit Indian Statistical Institute India

At present, adversarial attacks are designed in a task-specific fashion. However, for downstream computer vision tasks such as image captioning, image segmentation etc., the current deep learning systems use an image classifier like VGG16, ResNet50, Inception-v3 etc. as a feature extractor. Keeping this in mind, we propose Mimic and Fool, a task agnostic adversarial attack. Given a feature extractor, the proposed attack finds an adversarial image which can mimic the image feature of the original image. This ensures that the two images give the same (or similar) output regardless of the task. We randomly select 1000 MSCOCO validation images for experimentation. We perform experiments on two image captioning models, Show and Tell, Show Attend and Tell and one VQA model, namely, end-to-end neural module network (N2NMN). The proposed attack achieves success rate of 74.0%, 81.0% and 89.6% for Show and Tell, Show Attend and Tell and N2NMN respectively. We also propose a slight modification to our attack to generate natural-looking adversarial images. In addition, it is shown that the proposed attack also works for invertible architecture. Since Mimic and Fool only requires information about the feature extractor of the model, it can be considered as a gray-box attack. Copyright © 2019, The Authors. All rights reserved.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

An approach to a rough set decision system for classification of different heart diseases

An approach to a rough set decision system for classificatio...

引用

AMSE International Conference on Modelling and Simulation, MS'2004

作者： Mitra, S. Mitra, M. Chattopadhya, S. Chaudhuri, B.B. Department of Applied Physics Faculty of Technology University of Calcutta India Computer Vision and Pattern Recognition Unit Indian Statistical Institute Kolkata India JIS College of Engineering Kalyani India

In this paper a rule based rough set decision system for development of a disease inference engine is described. For this purpose an off-line data acquisition system of paper electrocardiogram (ECG) records are developed using image processing techniques. A QRS detector is developed for detection of R-R interval from ECG waves. After detection of this R-R interval the P and T waves are detected based on syntactic approaches and different time-plane features are extracted from every ECG signals. From a knowledgebase which is developed from the feedback of different reputed cardiologists and consultation of different medical books the essential time plane features for ECG interpretation have been selected. Finally, a rule-based roughest decision system is generated for the development of an inference engine for disease identification from these time-plane features.

关键词： Inference engines

来源：评论

学校读者我要写书评

暂无评论

Skew estimation and correction for form documents using wavelet decomposition

引用

2nd International Conference on Image Analysis and Recognition, ICIAR 2005

作者： Xi, Dihua Kamel, Mohamed Lee, Seong-Whan Pattern Analysis and Machine Intelligence Research Group Department of Electrical and Computer Engineering University of Waterloo Waterloo Canada Center for Artificial Vision Research Korea University Seoul Korea Republic of

ISBN: (纸本)3540290699

Form document image processing has become an increasingly essential technology in office automation tasks. One of the problems is that the document image may appear skewed for many reasons. Therefore, the skew estimation plays an important role in any automatic document analysis system. In the past few years, many algorithms have been developed to detect the skew angle of text document images. However, these algorithms suffer from two major deficiencies. Firstly, most of them suppose that the original image is monochrome and therefore they are not suitable to apply to documents with a complicated background. Secondly, most of the current methods were developed for general document images that are not as complicated as form documents. In this paper, we present a new approach to skew detection for grey-level form document images. In our system, image decomposition by 2D wavelet transformations is used to estimate the skew angle. © Springer-Verlag Berlin Heidelberg 2005.

关键词： Image processing

来源：评论

学校读者我要写书评

暂无评论

A word association based approach for improving retrieval performance from noisy OCRed text 6

A word association based approach for improving retrieval pe...

引用

6th International Conference on Knowledge Discovery and Information Retrieval, KDIR 2014

作者： Chakraborty, Anirban Ghosh, Kripabandhu Roy, Utpal Computer Vision and Pattern Recognition Unit Indian Statistical Institute 203 B. T. Road Kolkata700108 India Department of Computer and System Sciences Siksha-Bhavana Visva-Bharati Santiniketan731235 India

ISBN: (纸本)9789897580482

OCR errors hurt retrieval performance to a great extent. Research has been done on modelling and correction of OCR errors. However, most of the existing systems use language dependent resources or training texts for studying the nature of errors. Not much research has been reported on improving retrieval performance from erroneous text when no training data is available. We propose an algorithm of detecting OCR errors and improving retrieval performance from the erroneous corpus. We present two versions of the algorithm: one based on word cooccurrence and the other based on Pointwise Mutual Information. Our algorithm does not use any training data or any language specific resources like thesaurus. It also does not use any knowledge about the language except that the word delimiter is a blank space. We have tested our algorithm on erroneous Bangla FIRE collection and obtained significant improvements. Copyright © 2014 SCITEPRESS - Science and Technology Publications All rights reserved.

关键词： Errors

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：