Defending Neural ODE Image Classifiers from Adversarial Attacks with Tolerance Randomization.
Fabio CarraraRoberto CaldelliFabrizio FalchiGiuseppe AmatoPublished in: ICPR Workshops (6) (2020)
Keyphrases
- input image
- image features
- single image
- image classification
- image data
- multiscale
- extracted features
- segmentation method
- image analysis
- image collections
- edge detection
- image content
- network architecture
- test images
- training data
- image representation
- segmentation algorithm
- feature points
- image denoising
- classifier training
- neural network
- feature set
- training set
- image segmentation
- gray level
- image regions
- low level
- high resolution
- ddos attacks
- decision trees