Redirigiendo al acceso original de articulo en 17 segundos...
Inicio  /  Applied Sciences  /  Vol: 11 Par: 11 (2021)  /  Artículo
ARTÍCULO
TITULO

Methods for Preventing Visual Attacks in Convolutional Neural Networks Based on Data Discard and Dimensionality Reduction

Nikita Andriyanov    

Resumen

The article is devoted to the study of convolutional neural network inference in the task of image processing under the influence of visual attacks. Attacks of four different types were considered: simple, involving the addition of white Gaussian noise, impulse action on one pixel of an image, and attacks that change brightness values within a rectangular area. MNIST and Kaggle dogs vs. cats datasets were chosen. Recognition characteristics were obtained for the accuracy, depending on the number of images subjected to attacks and the types of attacks used in the training. The study was based on well-known convolutional neural network architectures used in pattern recognition tasks, such as VGG-16 and Inception_v3. The dependencies of the recognition accuracy on the parameters of visual attacks were obtained. Original methods were proposed to prevent visual attacks. Such methods are based on the selection of ?incomprehensible? classes for the recognizer, and their subsequent correction based on neural network inference with reduced image sizes. As a result of applying these methods, gains in the accuracy metric by a factor of 1.3 were obtained after iteration by discarding incomprehensible images, and reducing the amount of uncertainty by 4?5% after iteration by applying the integration of the results of image analyses in reduced dimensions.

 Artículos similares

       
 
Sirui Shen, Daobin Zhang, Shuchao Li, Pengcheng Dong, Qing Liu, Xiaoyu Li and Zequn Zhang    
Heterogeneous graph neural networks (HGNNs) deliver the powerful capability to model many complex systems in real-world scenarios by embedding rich structural and semantic information of a heterogeneous graph into low-dimensional representations. However... ver más
Revista: Applied Sciences

 
Viktor Blagovechshenskiy, Akhmetkal Medeu, Tamara Gulyayeva, Vitaliy Zhdanov, Sandugash Ranova, Aidana Kamalbekova and Ulzhan Aldabergen    
The assessment and forecast of avalanche danger are very important means of preventing avalanche fatalities, especially in recreational areas. The use of artificial intelligence methods for these purposes significantly increases the accuracy of avalanche... ver más
Revista: Water

 
Tatjana Bakran-Petricioli, Dajana Kujund?ic, Martina Naranda, Donat Petricioli, Lucija Petricioli and Silvija Kipson    
In the last few years, the endemic Mediterranean bivalve Pinna nobilis has been exposed to dramatic mortality in its entire area, which could lead to the extinction of the species. Throughout the Mediterranean, a lot of effort is being put into finding w... ver más

 
Zhiwei Xiang, Zhenxing Gao, Jiming Liu and Yangyang Zhang    
Discovering and mitigating potential risks in advance is essential for preventing aviation accidents on routine flights. Although anomaly detection-based explanation techniques have successfully uncovered potential risks for proactive flight safety manag... ver más
Revista: Aerospace

 
Xinxi Lu, Lijuan Yuan, Ruifeng Li, Zhihuan Xing, Ning Yao and Yichun Yu    
In recent years, the development of computer technology has promoted the informatization and intelligentization of hospital management systems and thus produced a large amount of medical data. These medical data are valuable resources for research. We ca... ver más
Revista: Algorithms