A Simple Way to Make Neural Networks Robust Against Diverse Image Corruptions.
Evgenia RusakLukas SchottRoland S. ZimmermannJulian BitterwolfOliver BringmannMatthias BethgeWieland BrendelPublished in: ECCV (3) (2020)
Keyphrases
- neural network
- image noise
- single image
- input image
- image analysis
- image features
- template matching
- partial occlusion
- image data
- image classification
- image representation
- multiscale
- pattern recognition
- geometric distortions
- edge detection
- region of interest
- image segmentation
- image content
- test images
- image structure
- high resolution
- image collections
- salt pepper
- image watermarking
- image retrieval
- feature descriptors
- pixel values
- segmentation algorithm
- lighting conditions
- image regions
- digital images
- natural images
- image matching
- image sequences
- hough transform