检索结果-内蒙古大学图书馆

A Real-Time Multi-Stage Architecture for Pose Estimation of Zebrafish Head with Convolutional Neural Networks

Journal of Computer Science & Technology 2021年第2期36卷 434-444页

作者： Zhang-Jin Huang Xiang-Xiang He Fang-Jun Wang Qing Shen School of Computer Science and Technology University of Science and Technology of ChinaHefei 230027China School of Data Science University of Science and Technology of ChinaHefei 230027China Anhui Province Key Laboratory of Software in Computing and Communication Hefei 230027China

In order to conduct optical neurophysiology experiments on a freely swimming zebrafish,it is essential to quantify the zebrafish head to determine exact lighting *** efficiently quantify a zebrafish head's behaviors with limited resources,we propose a real-time multi-stage architecture based on convolutional neural networks for pose estimation of the zebrafish head on *** stage is implemented with a small neural ***,a light-weight object detector named Micro-YOLO is used to detect a coarse region of the zebrafish head in the first *** the second stage,a tiny bounding box refinement network is devised to produce a high-quality bounding box around the zebrafish ***,a small pose estimation network named tiny-hourglass is designed to detect keypoints in the zebrafish *** experimental results show that using Micro-YOLO combined with RegressNet to predict the zebrafish head region is not only more accurate but also much faster than Faster R-CNN which is the representative of two-stage *** with DeepLabCut,a state-of-the-art method to estimate poses for user-defined body parts,our multi-stage architecture can achieve a higher accuracy,and runs 19x faster than it on CPUs.

关键词： convolutional neural network pose estimation real-time zebrafish

来源：评论

学校读者我要写书评

暂无评论

Imperceptible Adversarial Attack on S Channel of HSV Colorspace

Imperceptible Adversarial Attack on S Channel of HSV Colorsp...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Tong Zhu Zhaoxia Yin Wanli Lyu Jiefei Zhang Bin Luo Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Anhui University Shanghai Key Laboratory of Multidimensional Information Processing East China Normal University Shanghai China School of Communication and Electronic Engineering East China Normal University Shanghai China

Deep neural network models are vulnerable to subtle but adversarial perturbations that alter the model. Adversarial perturbations are typically computed for RGB images and, therefore, are evenly distributed among RGB channels. Compared with RGB images, HSV images can express the Hue, saturation, and brightness more intuitively. We find that the adversarial perturbation in the S-channel ensures a high attack success rate, while the perturbation is small, and the visual quality of the adversarial examples is good. Using this finding, we propose an attack method, SPGD, to improve the visual quality of adversarial examples by generating perturbations on the S-channel. Based on the attack principle of the PGD method, the RGB image was converted into an HSV image. The gradient calculated by the model on the S channel was superimposed on the S channel and then combined with the non-interference H and V channels to convert back to the RGB image. The iteration stops until the attack succeed. We compare the SPGD method with the existing state-of-the-art attack methods. The results show that SPGD minimizes pixel perturbation while maintaining a high attack success rate and achieves the best results in terms of structural similarity, imperceptibility, the minimum number of iterations, and the shortest run time.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Scaling Up Single Image Dehazing Algorithm by Cross-Data Vision Alignment for Richer Representation Learning and Beyond

引用

IEEE Transactions on Instrumentation and Measurement 2025年 74卷

作者： Shi, Yukai Weng, Zhipeng Lin, Yupei Shi, Cidan Guo, Xingyuan Yang, Xiaojun Lin, Liang Guangdong University of Technology School of Information Engineering Guangzhou510006 China Anhui University Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Hefei230601 China Southern Power Grid Ltd. Guangzhou510000 China Ministry of Education of China Key Laboratory of Photonic Technology for Integrated Sensing and Communication Guangdong Guangzhou510006 China Sun Yat-sen University School of Computer Science Guangzhou510006 China

In recent years, deep neural network tasks have increasingly relied on high-quality image inputs. With the development of high-resolution representation learning, the task of image dehazing has received significant attention. Previously, many methods collected diverse image data for large-scale training to boost the performance on a target scene. Ignoring the domain gap between different data, former dehazing methods simply adopt multiple datasets for explicit large-scale training, which often makes the methods themselves violated. To address this problem, we propose a novel method of cross-data vision alignment for richer representation learning to improve the existing dehazing methodology. Specifically, we call for the internal and external knowledge should be further adapted in a self-supervised manner to fill up the domain gap. By using cross-data external alignment, the datasets inherit samples from different domains that are firmly aligned, making the model learn more robust and generalizable features. By using the internal augmentation method, the model can fully exploit local information within the images and then obtain more image details. To demonstrate the effectiveness of our proposed method, we conduct training on the natural image dataset (NID). Experimental results show that our method clearly resolves the domain gap in different dehazing datasets and presents a new pipeline for large-scale training in the dehazing task. Our approach significantly outperforms other advanced methods in dehazing and produces dehazed images that are closest to real haze-free images. © 1963-2012 IEEE.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Identification of Microservices through Processed Dynamic Traces and Static Calls

Identification of Microservices through Processed Dynamic Tr...

引用

Computer Science and Management Technology (ICCSMT), International Conference on

作者： Sijie Wu Cheng Zhang Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Anhui University School of Computer Science and Technology Anhui University State Key Lab for Novel Software Technology Nanjing University Hefei China

As the complexity of software systems continues to increase, traditional monolithic software architectures will become difficult to scale and maintain. Microservices as an emerging software architecture it advocates decentralized management, service autonomy, and distinct module boundaries. These features of microservices make software systems easy to maintain, deploy, and modify. However, microservice identification is still not well addressed due to the lack of an effective approach. The existing work on microservice identification mainly starts from the system functionality perspective and does not take into account the actual business rules, which leads to microservice candidates not conforming to the concept of microservices. To address this challenge, we propose an automated approach to microservice identification that combines dynamic and static analysis and takes into account the actual business. Our proposed approach starts with a dynamic and static analysis of the monolithic system. Once the behavioral data of the system is collected, the behavioral data will be further processed according to our proposed guidelines. Then, the proposed approach calculates the degree of the correlation between classes in a monolithic system. Finally, we propose an improved clustering algorithm based on K-means to classify the classes in a monolithic system into different microservices. The proposed approach is validated by comparison with other approaches on open source projects.

关键词：

来源：评论

学校读者我要写书评

暂无评论

FTG: Score-based black-box watermarking by fragile trigger generation for deep model integrity verification

引用

Journal of Information and Intelligence 2024年第1期2卷 28-41页

作者： Heng Yin Zhaoxia Yin Zhenzhe Gao Hang Su Xinpeng Zhang Bin Luo Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Anhui UniversityHefei 230000China Shanghai key Laboratory of Multidimensional Information Processing East China Normal UniversityShanghai 200241China School of Communication and Electronic Engineering East China Normal UniversityShanghai 200241China Department of Computer Science and Technology Tsinghua UniversityBeijing 100000China School of Computer Science and Technology Fudan UniversityShanghai 200241China

Deep neural networks(DNNs)are widely used in real-world applications,thanks to their exceptional performance in image ***,their vulnerability to attacks,such as Trojan and data poison,can compromise the integrity and stability of DNN ***,it is crucial to verify the integrity of DNN models to ensure their *** research on model watermarking for integrity detection has encountered the issue of overexposure of model parameters during embedding and extraction of the *** address this problem,we propose a novel score-based black-box DNN fragile watermarking framework called fragile trigger generation(FTG).The FTG framework only requires the prediction probability distribution of the final output of the classifier during the watermarking *** generates different fragile samples as the trigger,based on the classification prediction probability of the target classifier and a specified prediction probability mask to watermark *** prediction probability masks can promote the generation of fragile samples in corresponding distribution *** whole watermarking process does not affect the performance of the target *** verifying the watermarking information,the FTG only needs to compare the prediction results of the model on the samples with the previous *** a result,the required model parameter information is reduced,and the FTG only needs a few samples to detect slight modifications in the *** results demonstrate the effectiveness of our proposed method and show its superiority over related *** FTG framework provides a robust solution for verifying the integrity of DNN models,and its effectiveness in detecting slight modifications makes it a valuable tool for ensuring the security and stability of DNN applications.

关键词： Neural network Sensitive sample Fragile watermarking Model integrity Fragile trigger

来源：评论

学校读者我要写书评

暂无评论

Reversible Data Hiding Based on Automatic Contrast Enhancement Using Histogram Expansion

SSRN

引用

SSRN 2022年

作者： Lyu, WanLi Yue, YaJie Yin, Zhaoxia Anhui Province Key Laboratory of Multimodal Cognitive Computation School of Computer Science and Technology Anhui Medical University 230601 China Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Anhui Medical University 230601 China School of Communication & Electronic Engineering East China Normal University Shanghai200241 China

This paper proposes a novel ACERDH method using histogram expansion. Considering the local properties of the histogram, when an appropriate range rather than a global location is selected, the contrast of the whole image is improved. Meanwhile, the global average brightness is used as a reference to determine the shift direction to prevent the image from being over-enhanced. Moreover, the contrast can be improved adaptively when a reasonable number of data is embedded at selected embedding points. Experimental results show that compared with state-of-the-art methods, when the payloads are the same, the proposed method achieves better contrast and maintains good visual quality. © 2022, The Authors. All rights reserved.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Adversarial Example Defense via Perturbation Grading Strategy

arXiv

引用

arXiv 2022年

作者： Zhu, Shaowei Lyu, Wanli Li, Bin Yin, Zhaoxia Luo, Bin Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Anhui University Hefei China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen Key Laboratory of Media Security Shenzhen University Shenzhen China School of Communication & Electronic Engineering East China Normal University Shanghai China

Deep Neural Networks have been widely used in many fields. However, studies have shown that DNNs are easily attacked by adversarial examples, which have tiny perturbations and greatly mislead the correct judgment of DNNs. Furthermore, even if malicious attackers cannot obtain all the underlying model parameters, they can use adversarial examples to attack various DNN-based task systems. Researchers have proposed various defense methods to protect DNNs, such as reducing the aggressiveness of adversarial examples by preprocessing or improving the robustness of the model by adding modules. However, some defense methods are only effective for small-scale examples or small perturbations but have limited defense effects for adversarial examples with large perturbations. This paper assigns different defense strategies to adversarial perturbations of different strengths by grading the perturbations on the input examples. Experimental results show that the proposed method effectively improves defense performance. In addition, the proposed method does not modify any task model, which can be used as a preprocessing module, which significantly reduces the deployment cost in practical applications. Copyright © 2022, The Authors. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Robust image steganography against lossy JPEG compression based on embedding domain selection and adaptive error correction

arXiv

引用

arXiv 2023年

作者： Duan, Xiaolong Li, Bin Yin, Zhaoxia Zhang, Xinpeng Luo, Bin Anhui Province Key Laboratory of Multimodal Cognitive Computation School of Computer Science and Technology Anhui University Hefei230601 China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen Key Laboratory of Media Security Shenzhen University Shenzhen518060 China School of Communication & Electronic Engineering East China Normal University Shanghai200241 China School of Computer Science Fudan University Shanghai200433 China

Transmitting images for communication on social networks has become routine, which is helpful for covert communication. The traditional steganography algorithm is unable to successfully convey secret information since the social network channel will perform lossy operations on images, such as JPEG compression. Previous studies tried to solve this problem by enhancing the robustness or making the cover adapt to the channel processing. In this study, we proposed a robust image steganography method against lossy JPEG compression based on embedding domain selection and adaptive error correction. To improve anti-steganalysis performance, the embedding domain is selected adaptively. To increase robustness and lessen the impact on anti-steganalysis performance, the error correction capacity of the error correction code is adaptively adjusted to eliminate redundancy. The experimental results show that the proposed method achieves better anti-steganalysis and robustness. Copyright © 2023, The Authors. All rights reserved.

关键词： Error correction

来源：评论

学校读者我要写书评

暂无评论

High Capacity Reversible Data Hiding for Encrypted 3D Mesh Models Based on Topology

arXiv

引用

arXiv 2022年

作者： Tang, Yun Cheng, Lulu Lyv, Wanli Yin, Zhaoxia Anhui Province Key Laboratory of Multimodal Cognitive Computation School of Computer Science and Technology Anhui University 230601 China School of Communication & Electronic Engineering East China Normal University Shanghai200241 China

Reversible data hiding in encrypted domain(RDH-ED) can not only protect the privacy of 3D mesh models and embed additional data, but also recover original models and extract additional data losslessly. However, due to the insufficient use of model topology, the existing methods have not achieved satisfactory results in terms of embedding capacity. To further improve the capacity, a RDH-ED method is proposed based on the topology of the 3D mesh models, which divides the vertices into two parts: embedding set and prediction set. And after integer mapping, the embedding ability of the embedding set is calculated by the prediction set. It is then passed to the data hider for embedding additional data. Finally, the additional data and the original models can be extracted and recovered respectively by the receiver with the correct keys. Experiments declare that compared with the existing methods, this method can obtain the highest embedding capacity. Copyright © 2022, The Authors. All rights reserved.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

High-Capacity Reversible Data Hiding in Encrypted 3d Mesh Models Based on Multi-Msb Prediction

SSRN

引用

SSRN 2022年

作者： Lyu, Wan-Li Cheng, Lulu Yin, Zhaoxia Anhui Province Key Laboratory of Multimodal Cognitive Computation School of Computer Science and Technology Anhui University 230601 China School of Communication & Electronic Engineering East China Normal University Shanghai200241 China

The interaction of massive data in the network world poses a enormous threat to data security. Reversible data hiding in encrypted domain (RDH-ED) technology ensures the security of data storage. With the advancement of multimedia technology, three-dimensional (3D) models are widely used on the network. Recently, 3D models have also been used by researchers as cover mediums for RDH-ED technology. However, it is still challenging to improve the embedding rate under the premise of ensuring reversibility. This paper proposes an RDH-ED method to improve the embedding rate significantly, and the 3D mesh model can be restored after data extraction. Firstly, the vertices of the 3D mesh models are classified into "embedded set" and prediction set" based on odd-even property of indices. Based on the high correlation of the two sets, multiple most significant bit (multi-MSB) prediction is adopted to calculate room redundancy. Arithmetic coding compresses auxiliary information generated during the reserved room phase to further free up space. In addition, the correlation between two sets is used to recover vertices that were modified due to data hiding. Experimental results show the proposed method greatly improves the embedding rate compared with state-of-the-art methods and guarantees model recovery with high quality. © 2022, The Authors. All rights reserved.

关键词： Forecasting

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：