检索结果-内蒙古大学图书馆

Optimal KAZE and AKAZE Features for Facial Similarity Matching 1

Proceedings of the 7th International Conference on Advances in Computing and Data Sciences, ICACDS 2023

作者： Vinay, A. Vasu, Kishan Athirala Lodha, Pranav Yogi Natarajan, S. Sudarshan, T.S.B. Center for Pattern Recognition and Machine Intelligence PES University Bengaluru560085 India

ISBN: (数字)9783031379406

ISBN: (纸本)9783031379390

Face recognition is one of the premier disciplines in the vast field of computer vision and image analysis. A popular method is the Gaussian scale space analysis which limits the performance by smoothing both the noise and natural boundaries in the same proportion. In order to prevent the loss of natural boundaries to smoothing we tap into nonlinear scale space techniques such as KAZE and Accelerated KAZE. KAZE is a multistage 2-D feature detection and description algorithm. It makes use of AOS schemes to develop the nonlinear scale space for analysis. Though the results are satisfactory, it is computationally intense as they solve a humongous module of linear equations. To ascertain the mentioned limitation of KAZE we make use of Accelerated-KAZE, which uses pyramidal structure with Fast Explicit Diffusion incorporated in it, thus minimizing the computation in the step of feature detection of nonlinear scale space. Also, with the use of M-LDB (Modified-Local Difference Binary) descriptor the problem of rotation is solved. Usage of RANSAC after processing in two methods had some disadvantages. It gave bad results when the inliers ratio in the dataset is low. Thus, Optimal RANSAC is employed which works well even when the inliers ratio is as low as 5%. The proposed methods are tested on many standard datasets and various performance parameters. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

Discrete Cross-Modal Alignment Enables Zero-Shot Speech Translation

Discrete Cross-Modal Alignment Enables Zero-Shot Speech Tran...

引用

2022 Conference on Empirical Methods in Natural Language Processing, EMNLP 2022

作者： Wang, Chen Liu, Yuchen Chen, Boxing Zhang, Jiajun Luo, Wei Huang, Zhongqiang Zong, Chengqing National Laboratory of Pattern Recognition Institute of Automation CAS Beijing China Machine Intelligence Technology Lab Alibaba DAMO Academy China

End-to-end Speech Translation (ST) aims at translating the source language speech into target language text without generating the intermediate transcriptions. However, the training of end-to-end methods relies on parallel ST data, which are difficult and expensive to obtain. Fortunately, the supervised data for automatic speech recognition (ASR) and machine translation (MT) are usually more accessible, making zero-shot speech translation a potential direction. Existing zero-shot methods fail to align the two modalities of speech and text into a shared semantic space, resulting in much worse performance compared to the supervised ST methods. In order to enable zero-shot ST, we propose a novel Discrete Cross-Modal Alignment (DCMA) method that employs a shared discrete vocabulary space to accommodate and match both modalities of speech and text. Specifically, we introduce a vector quantization module to discretize the continuous representations of speech and text into a finite set of virtual tokens, and use ASR data to map corresponding speech and text to the same virtual token in a shared codebook. This way, source language speech can be embedded in the same semantic space as the source language text, which can be then transformed into target language text with an MT module. Experiments on multiple language pairs demonstrate that our zero-shot ST method significantly improves the SOTA, and even performs on par with the strong supervised ST baselines. © 2022 Association for Computational Linguistics.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Enhancement of Degraded CCTV Footage for Forensic Analysis 3rd

Enhancement of Degraded CCTV Footage for Forensic Analysis

引用

3rd International Conference on Innovative Computing and Communication, ICICC 2020

作者： Vinay, A. Lokesh, Aditya Kamath, Vinayaka R. Murty, K.N.B. Natarajan, S. Center for Pattern Recognition and Machine Intelligence PES University Bengaluru India

ISBN: (纸本)9789811551123

Forensic analysis has proved to be one of the most utilitarian tool in investigating crime. Forensic analysis provides evidence/basic information of the said crime through analysis of physical evidence. In this paper, we present a scintillating technique to enhance the image quality of CCTV video to assist in the investigation of criminal cases. This paper, address the issue of lack of a budget surplus for a high-quality image/video capturing device. The problem of enhancing images is addressed by pure image processing methods and machine learning techniques. The paper analyzes both techniques and further concludes that the machine learning approach produces a more efficient result. The applications of this technique can range from simple image filtering/beautification to forensic image processing. © 2021, Springer Nature Singapore Pte Ltd.

关键词： Crime

来源：评论

学校读者我要写书评

暂无评论

CMQA: A Dataset of Conditional Question Answering with Multiple-Span Answers 29

CMQA: A Dataset of Conditional Question Answering with Multi...

引用

29th International Conference on Computational Linguistics, COLING 2022

作者： Ju, Yiming Wang, Weikang Zhang, Yuanzhe Zheng, Suncong Liu, Kang Zhao, Jun National Laboratory of Pattern Recognition Institute of Automation CAS Beijing China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing China Machine Learning Platform Department Tencent China

Forcing the answer of the Question Answering (QA) task to be a single text span might be restrictive since the answer can be multiple spans in the context. Moreover, we found that multi-span answers often appear with two characteristics when building the QA system for a real-world application. First, multi-span answers might be caused by users lacking domain knowledge and asking ambiguous questions, which makes the question need to be answered with conditions. Second, there might be hierarchical relations among multiple answer spans. Some recent span-extraction QA datasets include multi-span samples, but they only contain unconditional and parallel answers, which cannot be used to tackle this problem. To bridge the gap, we propose a new task: conditional question answering with hierarchical multi-span answers, where both the hierarchical relations and the conditions need to be extracted. Correspondingly, we introduce CMQA, a Conditional Multiple-span Chinese Question Answering dataset to study the new proposed task. The final release of CMQA consists of 7,861 QA pairs and 113,089 labels, where all samples contain multi-span answers, 50.4% of samples are conditional, and 56.6% of samples are hierarchical. CMQA can serve as a benchmark to study the new proposed task and help study building QA systems for real-world applications. The low performance of models drawn from related literature shows that the new proposed task is challenging for the community to solve. CMQA can be accessed at https://***/juyiming/CMQA. © 2022 Proceedings - International Conference on Computational Linguistics, COLING. All rights reserved.

关键词： Domain Knowledge

来源：评论

学校读者我要写书评

暂无评论

Shop Signboards Detection Using the ShoS Dataset 3rd

Shop Signboards Detection Using the ShoS Dataset

引用

3rd International Conference on pattern recognition and Artificial intelligence, ICPRAI 2022

作者： Almuhajri, Mrouj Suen, Ching Y. Centre for Pattern Recognition and Machine Intelligence Concordia University MontrealQCH3G 2W1 Canada Department of Computer Science Saudi Electronic University Riyadh Saudi Arabia

ISBN: (纸本)9783031092817

This work was driven by the possibility of enabling beneficial applications for humans and municipal agencies utilizing deep learning techniques to detect shop signboards. Data is the foundation of successful AI-based applications. Despite the abundance of data, providing datasets for such applications is a challenge as they have to be refined and characterized accurately for better performance. This paper describes the process of collecting the ShoS dataset from Google Street Views for the detection of the signboards of shops on streets. A total of 10k storefront signboards were captured and fully annotated within 7500 images. The ShoS datasets were tested by running different baseline methodology using one-stage detectors YOLO and SSD. Out of the tested methods and variances, the YOLOv3 yields the highest performance reaching 94.23% (mAP@0.5). © 2022, Springer Nature Switzerland AG.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Multi-label learning on principles of reverse k-nearest neighbourhood

Multi-label learning on principles of reverse k-nearest neig...

引用

作者： Sadhukhan, Payel Palit, Sarbani Machine Intelligence Unit Indian Statistical Institute Kolkata India Computer Vision and Pattern Recognition Unit Indian Statistical Institute Kolkata India

In this article, we present a novel neighbourhood based multi-label classifier, Multi-label Learning on principles of Reverse k-Nearest Neighbourhood (ML-RkNN) where we estimate the neighbourhood of the points on the basis of their reverse k-nearest neighbourhood (RkNN). Through RkNN, for the same value of k, we get different number of neighbours for different instances and this happens adaptively according to the neighbourhood configuration of the points. The automatically adaptive neighbourhood helps us in better learning of the local configurations around the points. Our scheme also facilitates implicit handling of the local imbalances prevailing in the datasets by comparing the class distributions of the test points and their reverse nearest neighbours. This implicit and adaptive handling is particularly useful for multi-label label datasets, whose labels are differentially imbalanced. Empirical study is performed on 10 real-world multi-label datasets considering five neighbourhood based multi-label learners. Macro F1 is used as the evaluating metric. The proposed method has given statistically superior and statistically comparable performances with respect to three and two comparing methods respectively. Additionally, we have explored the use of two different distance metrics, Euclidean and Jaccard in our scheme for nominal datasets. © 2020 John Wiley & Sons, Ltd.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

CCLSL: Combination of Contrastive Learning and Supervised Learning for Handwritten Mathematical Expression recognition 16th

CCLSL: Combination of Contrastive Learning and Supervised ...

引用

16th Asian Conference on Computer Vision, ACCV 2022

作者： Lin, Qiqiang Huang, Xiaonan Bi, Ning Suen, Ching Y. Tan, Jun School of Mathematics and Computational Science Sun Yat-Sen University Guangzhou510275 China Guangdong Province Key Laboratory of Computational Science Sun Yat-Sen University Guangzhou510275 China Centre for Pattern Recognition and Machine Intelligence Concordia University MontrealQCH3G 1M8 Canada

ISBN: (纸本)9783031262838

Handwritten Mathematical Expressions differ considerably from ordinary linear handwritten texts, due to their two-dimentional structures plus many special symbols and characters. Hence, HMER(Handwritten Mathematical Expression recognition) is a lot more challenging compared with normal handwriting recognition. At present, the mainstream offline recognition systems are generally built on deep learning methods, but these methods can hardly cope with HEMR due to the lack of training data. In this paper, we propose an encoder-decoder method combining contrastive learning and supervised learning(CCLSL), whose encoder is trained to learn semantic-invariant features between printed and handwritten characters effectively. CCLSL improves the robustness of the model in handwritten styles. Extensive experiments on CROHME benchmark show that without data enhancement, our model achieves an expression accuracy of 58.07% on CROHME2014, 55.88% on CROHME2016 and 59.63% on CROHME2019, which is much better than all previous state-of-the-art methods. Furthermore, our ensemble model added a boost of 2.5% to 3.4% to the accuracy, achieving the state-of-the-art performance on public CROHME datasets for the first time. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Advancing biological super-resolution microscopy through deep learning: a brief review

引用

Biophysics Reports 2021年第4期7卷 253-266页

作者： Tianjie Yang Yaoru Luo Wei Ji Ge Yang Institute of Biophysics Chinese Academy of Sciences College of Life Sciences University of Chinese Academy of Sciences Laboratory of Computational Biology and Machine Intelligence School of Artificial IntelligenceUniversity of Chinese Academy of Sciences National Laboratory of Pattern Recognition Institute of AutomationChinese Academy of Sciences

Biological super-resolution microscopy is a new generation of imaging techniques that overcome the ～200 nm diffraction limit of conventional light microscopy in spatial resolution. By providing novel spatial or spatiotemporal information on biological processes at nanometer resolution with molecular specificity, it plays an increasingly important role in biomedical sciences. However, its technical constraints also require trade-offs to balance its spatial resolution, temporal resolution, and light exposure of samples. Recently, deep learning has achieved breakthrough performance in many image processing and computer vision tasks. It has also shown great promise in pushing the performance envelope of biological super-resolution microscopy. In this brief review, we survey recent advances in using deep learning to enhance the performance of biological super-resolution microscopy, focusing primarily on computational reconstruction of super-resolution images. Related key technical challenges are discussed. Despite the challenges, deep learning is expected to play an important role in the development of biological super-resolution microscopy. We conclude with an outlook into the future of this new research area.

关键词： Super-resolution microscopy Image super-resolution Deep learning Image reconstruction Fluorescence microscopy

来源：评论

学校读者我要写书评

暂无评论

Residual and Ensemble Learning on Locally Derived Features for Face recognition 1

引用

5th International Symposium on Intelligent Systems Technologies and Applications, ISTA 2019

作者： Vinay, A. Gupta, Abhijay Bharadwaj, Aprameya Srinivasan, Arvind Balasubramanya Murthy, K.N. Natarajan, S. Centre for Pattern Recognition and Machine Intelligence PES University BengaluruKA560085 India

ISBN: (数字)9789811539145

ISBN: (纸本)9789811539138

In this paper, an efficient approach toward face recognition is presented. The proposed model is invariant to pose, expression, scale, illumination, and translation with the application of different techniques and implementation of their algorithms. The images are first preprocessed using different techniques like CS-LBP, XCS-LBP, CS-LDMP, etc. to remove noise and make it illumination invariant. In one implementation, these are fed as input to different descriptors. The keypoints detected are invariant to lighting condition. A wide variety of descriptors are applied to the detected keypoints in order to compare the quality of features computed. These descriptors are quantized to a single vector, representing the most salient features of the facial image. The vectors are then fed into the stacking classifier. In the other method, these preprocessed images are directly fed as input into the residual network. A comparison of the results by these two pipelines is studied on two benchmark face datasets, namely FACES94 and Grimace. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

Discrete Cross-Modal Alignment Enables Zero-Shot Speech Translation

arXiv

引用

arXiv 2022年

End-to-end Speech Translation (ST) aims at translating the source language speech into target language text without generating the intermediate transcriptions. However, the training of end-to-end methods relies on parallel ST data, which are difficult and expensive to obtain. Fortunately, the supervised data for automatic speech recognition (ASR) and machine translation (MT) are usually more accessible, making zero-shot speech translation a potential direction. Existing zero-shot methods fail to align the two modalities of speech and text into a shared semantic space, resulting in much worse performance compared to the supervised ST methods. In order to enable zero-shot ST, we propose a novel Discrete Cross-Modal Alignment (DCMA) method that employs a shared discrete vocabulary space to accommodate and match both modalities of speech and text. Specifically, we introduce a vector quantization module to discretize the continuous representations of speech and text into a finite set of virtual tokens, and use ASR data to map corresponding speech and text to the same virtual token in a shared codebook. This way, source language speech can be embedded in the same semantic space as the source language text, which can be then transformed into target language text with an MT module. Experiments on multiple language pairs demonstrate that our zero-shot ST method significantly improves the SOTA, and even performs on par with the strong supervised ST baselines1 Copyright © 2022, The Authors. All rights reserved.

关键词： Speech recognition

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：