Redirigiendo al acceso original de articulo en 21 segundos...
Inicio  /  Applied Sciences  /  Vol: 12 Par: 18 (2022)  /  Artículo
ARTÍCULO
TITULO

Stylized Pairing for Robust Adversarial Defense

Dejian Guan    
Wentao Zhao and Xiao Liu    

Resumen

Recent studies show that deep neural networks (DNNs)-based object recognition algorithms overly rely on object textures rather than global object shapes, and DNNs are also vulnerable to human-less perceptible adversarial perturbations. Based on these two phenomenons, we conjecture that the preference of DNNs on exploiting object textures for decisions is one of the most important reasons for the existence of adversarial examples. At present, most adversarial defense methods are directly related to adversarial perturbations. In this paper, we propose an adversarial defense method independent of adversarial perturbations, which utilizes a stylized pairing technique to encourage logits for a pair of images and the corresponding stylized image to be similar. With stylized pairing training, DNNs can better learn shape-biased representation. We have empirically evaluated the performance of our method through extensive experiments on CIFAR10, CIFAR100, and ImageNet datasets. Results show that the models with stylized pairing training can significantly improve their performance against adversarial examples.

 Artículos similares

       
 
Ioannis Polymenis, Maryam Haroutunian, Rose Norman and David Trodden    
Underwater Vehicles have become more sophisticated, driven by the off-shore sector and the scientific community?s rapid advancements in underwater operations. Notably, many underwater tasks, including the assessment of subsea infrastructure, are performe... ver más

 
Kamran Javed, Nizam Ud Din, Ghulam Hussain and Tahir Farooq    
Face photographs taken on a bright sunny day or in floodlight contain unnecessary shadows of objects on the face. Most previous works deal with removing shadow from scene images and struggle with doing so for facial images. Faces have a complex semantic ... ver más
Revista: Applied Sciences

 
Jiazhu Dai and Siwei Xiong    
Capsule networks are a type of neural network that use the spatial relationship between features to classify images. By capturing the poses and relative positions between features, this network is better able to recognize affine transformation and surpas... ver más
Revista: Algorithms

 
He Wang, Hua Zou and Dengyi Zhang    
Shadow removal is a fundamental task that aims at restoring dark areas in an image where the light source is blocked by an opaque object, to improve the visibility of shadowed areas. Existing shadow removal methods have developed for decades and yielded ... ver más
Revista: Information

 
Shuai Dong, Wei Wang, Wensheng Li and Kun Zou    
A 2D floor plan (FP) often contains structural, decorative, and functional elements and annotations. Vectorization of floor plans (VFP) is an object detection task that involves the localization and recognition of different structural primitives in 2D FP... ver más
Revista: Information