On the human-recognizability phenomenon of adversarially trained deep image classifiers.
Jonathan HellandNathan M. VanHoudnosPublished in: CoRR (2021)
Keyphrases
- image features
- image classification
- single image
- image retrieval
- input image
- image data
- image content
- multiscale
- training set
- pixel values
- image analysis
- image representation
- image pixels
- low level
- training data
- similarity measure
- image collections
- test images
- extracted features
- static images
- decision trees
- classifier training
- pre trained
- region of interest
- adaboost classifier
- image regions
- edge detection
- image segmentation
- template matching
- support vector machine svm
- visual features
- feature points
- feature set
- high resolution
- neural network