Targeted Attack of Deep Hashing Via Prototype-Supervised Adversarial Networks

Zhang, Zheng; Wang, Xunguang; Lu, Yao; Shen, Fumin; Zhu, Lei

doi:10.1109/tmm.2021.3097506

Cited by 10 publications

(4 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By simulating the diffusion dynamics of watercolor pigments on a digital canvas, the proposed method generates realistic watercolor-style paintings with desirable textures and color blending. In [22] focused on watercolor painting style transfer while maintaining temporal consistency in video sequences. The authors propose an approach that incorporates both spatial and temporal constraints to ensure consistent watercolor style transfer across frames, enabling the creation of visually coherent watercolor-style videos.…”

Section: Literature Surveymentioning

confidence: 99%

A New Automatic Watercolour Painting Algorithm Based on Dual Stream Image Segmentation Model with Colour Space Estimation

Zhou,

Sirivesmas,

Joneurairatana

et al. 2023

IJRITCC

View full text Add to dashboard Cite

Image processing plays a crucial role in automatic watercolor painting by manipulating the digital image to achieve the desired watercolor effect. segmentation in automatic watercolor painting algorithms is essential for region-based processing, color mixing and blending, capturing brushwork and texture, and providing artistic control over the final result. It allows for more realistic and expressive watercolor-like paintings by processing different image regions individually and applying appropriate effects to each segment. Hence, this paper proposed an effective Dual Stream Exception Maximization (DSEM) for automatic image segmentation. DSEM combines both color and texture information to segment an image into meaningful regions. This approach begins by converting the image from the RGB color space to a perceptually-based color space, such as CIELAB, to account for variations in lighting conditions and human perception of color. With the color space conversion, DSEM extracts relevant features from the image. Color features are computed based on the values of the color channels in the chosen color space, capturing the nuances of color distribution within the image. Simultaneously, texture features are derived by computing statistical measures such as local variance or co-occurrence matrices, capturing the textural characteristics of the image. Finally, the model is applied over the deep learning model for the classification of the color space in the painting. Simulation analysis is performed compared with conventional segmentation techniques such a CNN and RNN. The comparative analysis states that the proposed DSEM exhibits superior performance compared to conventional techniques in terms of color space estimation, texture analysis and region merging. The performance of classification with DSEM is ~12% higher than the conventional techniques.

show abstract

Section: Literature Surveymentioning

confidence: 99%

A New Automatic Watercolour Painting Algorithm Based on Dual Stream Image Segmentation Model with Colour Space Estimation

Zhou,

Sirivesmas,

Joneurairatana

et al. 2023

IJRITCC

View full text Add to dashboard Cite

show abstract

“…The traditional machine learning algorithms [24–27] for breast mass classification tasks on mammography need to be designed manually, which is difficult to fully exploit the features from breast mammography to obtain a satisfactory classification performance. The Convolutional Neural Network (CNN) can automatically extract features from image [28] and avoid designing the feature extraction methods manually, which is widely used in the study of breast cancer diagnosis on mammography [29, 30]. Gardezi et al.…”

Section: Introductionmentioning

confidence: 99%

“…The traditional machine learning algorithms [24][25][26][27] for breast mass classification tasks on mammography need to be designed manually, which is difficult to fully exploit the features from breast mammography to obtain a satisfactory classification performance. The Convolutional Neural Network (CNN) can automatically extract features from image [28] and avoid designing the feature extraction methods manually, which is widely used in the study of breast cancer diagnosis on mammography [29,30]. Gardezi et al [31], Qiu et al [32] and Jaffar et al [33] extracted the features from breast mammography by using a CNN, then obtained the diagnosis by classifying these extracted features using a classifier, such as SVM, logistic classifier and k-NN.…”

Section: Introductionmentioning

confidence: 99%

Breast mass classification based on supervised contrastive learning and multi‐view consistency penalty on mammography

et al. 2022

Self Cite

View full text Add to dashboard Cite

Breast cancer accounts for the largest number of patients among all cancers in the world. Intervention treatment for early breast cancer can dramatically extend a woman's 5‐year survival rate. However, the lack of public available breast mammography databases in the field of Computer‐aided Diagnosis and the insufficient feature extraction ability from breast mammography limit the diagnostic performance of breast cancer. In this paper, A novel classification algorithm based on Convolutional Neural Network (CNN) is proposed to improve the diagnostic performance for breast cancer on mammography. A multi‐view network is designed to extract the complementary information between the Craniocaudal (CC) and Mediolateral Oblique (MLO) mammographic views of a breast mass. For the different predictions of the features extracted from the CC view and MLO view of the same breast mass, the proposed algorithm forces the network to extract the consistent features from the two views by the cross‐entropy function with an added consistent penalty term. To exploit the discriminative features from the insufficient mammographic images, the authors learnt an encoder in the classification model to learn the invariable representations from the mammographic breast mass by Supervised Contrastive Learning (SCL) to weaken the side effect of colour jitter and illumination of mammographic breast mass on image quality degradation. The experimental results of all the classification algorithms mentioned in this paper on Digital Database for Screening Mammography (DDSM) illustrate that the proposed algorithm greatly improves the classification performance and diagnostic speed of mammographic breast mass, which is of great significance for breast cancer diagnosis.

show abstract

“…Recent works [6,11,22,30,32] have found the vulnerability of DNNs to adversarial examples, which are carefully crafted to mislead DNNs by adding a human-imperceptible perturbation to clean samples. It threatens the security of DNNs and has therefore inspired more and more research [33][34][35]45] in recent years.…”

Section: Introductionmentioning

confidence: 99%

Adaptive Temporal Grouping for Black-box Adversarial Attacks on Videos

Wei

Chen

Zhang

et al. 2022

Proceedings of the 2022 International Conference on Multimedia Retrieval

View full text Add to dashboard Cite

Deep-learning based video models, which have remarkable performance on action recognition tasks, are recently proved to be vulnerable to adversarial samples, even those generated in the black-box setting. However, these black-box attack methods are insufficient to attack videos models in real-world applications due to the requirement of lots of queries. To this end, we propose to boost the efficiency of black-box attacks on video recognition models. Although videos carry rich temporal information, they include redundant spatial information from adjacent frames. This motivates us to introduce the adaptive temporal grouping (ATG) method, which groups video frames by the similarity of their features extracted from the ImageNet-pretrained image model. By selecting one key-frame from each group, ATG helps any black-box attack methods to optimize the adversarial perturbations over key-frames instead of all frames, where the estimated gradient of key-frame is shared with other frames in each group. To balance the efficiency and precision of estimated gradients, ATG adaptively adjusts the group number by the magnitude of the current perturbation and the current query number. Through extensive experiments on the HMDB-51 dataset and the UCF-101 dataset, we demonstrate that ATG can significantly reduce the number of queries by more than 10% for the targeted attack.

show abstract

Targeted Attack of Deep Hashing Via Prototype-Supervised Adversarial Networks

Cited by 10 publications

References 52 publications

A New Automatic Watercolour Painting Algorithm Based on Dual Stream Image Segmentation Model with Colour Space Estimation

A New Automatic Watercolour Painting Algorithm Based on Dual Stream Image Segmentation Model with Colour Space Estimation

Breast mass classification based on supervised contrastive learning and multi‐view consistency penalty on mammography

Adaptive Temporal Grouping for Black-box Adversarial Attacks on Videos

Contact Info

Product

Resources

About