Computer vision tasks, such as image classification, semantic segmentation, and super resolution, are broadly utilized in many applications. Recent studies revealed that machinelearning-based models for the computer ...
详细信息
ISBN:
(数字)9781665490627
ISBN:
(纸本)9781665490627
Computer vision tasks, such as image classification, semantic segmentation, and super resolution, are broadly utilized in many applications. Recent studies revealed that machinelearning-based models for the computer vision tasks are vulnerable to adversarial attacks. Since the adversarial attack can disturb the computer vision models in real-world systems, many countermeasures have been proposed against the adversarial attacks, such as denoising, resizing, and machinelearning-based super resolution model as a preprocessing. Recently, a prior work demonstrated that the super resolution model as a preprocessing can be vulnerable to the adversarial attack targeted to the preprocessing itself, only when the perturbation is inactive before the preprocessing. However, we also found that the perturbation before the preprocessing can be another serious threat if the super resolution model is used for a mitigation of adversarial attacks. In this paper, we propose Layered Adversary Generation (LAG) that generates the adversarial example by recursively injecting noises to clean image in white-box environment. We then show that LAG is effective to attack a semantic segmentation model even if the super resolution models with/without two countermeasures as auxiliary methods such as resizing and denoising are adopted to mitigate the adversarial attacks. Furthermore, we demonstrate that LAG is transferable across other super resolution models. Lastly, we discuss our attack method in gray-box and black-box environments, and suggests a mitigation for robust preprocessing.
Named Entity recognition (NER) is a stream of Natural language pro-cessing playing a pivotal role in the KDD process of datamining. the Last few years have seen a sudden surge in research about NER models, however, t...
详细信息
Face recognition technology, defined as the process of detecting and verifying individuals’ identities via image processing techniques in computer vision, has become increasingly vital in today’s technologically and...
详细信息
In recent years, the detection of targets at sea using computer vision methods has been improved. However, it is not enough to get information about the type of target. therefore, this paper uses a monocular vision wh...
详细信息
In current era, breast cancer is one of the serious issues for women as compared to all other cancers. Breast cancer occurs when cancer cells form in breast tissues. It can be fat or connective tissue in the breast an...
详细信息
Withthe rapid development of the Internet of things and artificial intelligence technologies, intelligent embedded systems face increasingly complex challenges and demands. this paper proposes an intelligent embedded...
详细信息
Structured Visual Content (SVC) such as graphs, flow charts, or the like are used by authors to illustrate various concepts. While such depictions allow the average reader to better understand the contents, images con...
详细信息
ISBN:
(数字)9781665490627
ISBN:
(纸本)9781665490627
Structured Visual Content (SVC) such as graphs, flow charts, or the like are used by authors to illustrate various concepts. While such depictions allow the average reader to better understand the contents, images containing SVCs are typically not machine-readable. this, in turn, not only hinders automated knowledge aggregation, but also the perception of displayed information for visually impaired people. In this work, we propose a synthetic dataset, containing SVCs in the form of images as well as ground truths. We show the usage of this dataset by an application that automatically extracts a graph representation from an SVC image. this is done by training a model via common supervised learning methods. As there currently exist no large-scale public datasets for the detailed analysis of SVC, we propose the Synthetic SVC (SSVC) dataset comprising 12,000 images with respective bounding box annotations and detailed graph representations. Our dataset enables the development of strong models for the interpretation of SVCs while skipping the time-consuming dense data annotation. We evaluate our model on both synthetic and manually annotated data and show the transferability of synthetic to real via various metrics, given the presented application. Here, we evaluate that this proof of concept is possible to some extend and lay down a solid baseline for this task. We discuss the limitations of our approach for further improvements. Our utilized metrics can be used as a tool for future comparisons in this domain. To enable further research on this task, the dataset is publicly available at https://***/3jN1pJJ.
Understanding the structure of multidimensional patterns, especially in unsupervised cases, is of fundamental importance in datamining, patternrecognition and machinelearning. Several algorithms have been proposed ...
详细信息
the Internet of things (IoT) smart devices IoT devices are widely used in power grid systems. Combining IoT with fingerprint recognition enhances device security, personalizes authentication, and improves user experie...
详细信息
Today, one of the emerging challenges faced by neurologists is to categorize Alzheimer's disease (AD). It is a type of neurodegenerative disorder and leads to progressive mental loss and is known as Alzheimer'...
详细信息
ISBN:
(纸本)9789819940707
Today, one of the emerging challenges faced by neurologists is to categorize Alzheimer's disease (AD). It is a type of neurodegenerative disorder and leads to progressive mental loss and is known as Alzheimer's disease (AD) (Tanveer et al. in Commun Appl 16:1–35, 2020). An immediate diagnosis of Alzheimer's disease is one of the requirements and developing an effective treatment strategy and stopping the disease’s progression. Resonance magnetic imaging (MRI) and CT scans can enable local changes in brain structure and quantify disease-related damage. the standard machinelearning algorithms are designed to detect AD to have poor performance because they were trained using insufficient sample data. In comparison with traditional machinelearning algorithms, deep learning models have shown superior performance in most of the research studies stated specific to diagnosis of AD. One of the elegant DL method is the convolutional neural network (CNN) and has helped to assist the early diagnosis of AD (Sethi et al. in BioMed Research international, 2022;Islam and Zhang in Proceedings IEEE/CVF 841 conference computing vision patternrecognition workshops (CVPRW), pp 1881–1883, 2018). However, in recent days advanced DL methods have also attempted for classification of AD, especially in MRI images (Tiwari et al. in Int J Nanomed 14:5541, 2019). the purpose of this paper is to propose a ResNet50 model for Alzheimer's disease, namely AD-ResNet50 for MRI images that incorporates two extensions known as transfer learning and SMOTE. this research uses the proposed method and compares it withthe standard deep models VGG19, InceptionResNet V2, and DenseNet169 with transfer learning and SMOTE (Chawla et al. in J Artif Intell Res 16:(1)321–357, 2002). the results demonstrate the efficiency of the proposed method, which outperforms the other three models tested. When compared with baseline deep learning models, the proposed model outperformed them in terms of accuracy and ROC values
暂无评论