Increasing the robustness of DNNs against image corruptions by playing the Game of Noise.
Evgenia RusakLukas SchottRoland S. ZimmermannJulian BitterwolfOliver BringmannMatthias BethgeWieland BrendelPublished in: CoRR (2020)
Keyphrases
- image noise
- image data
- input image
- image content
- image features
- background noise
- image retrieval
- image analysis
- additive noise
- game playing
- random noise
- image segmentation
- image structure
- multiscale
- image transformations
- computer games
- image classification
- single image
- geometric attacks
- salt pepper
- degraded images
- imperfect information
- test images
- high resolution
- pixel values
- low frequency
- board game
- noise level
- gaussian noise
- online game
- noise free
- imaging process
- pac man
- denoising methods
- denoising algorithm
- nash equilibrium
- image regions
- segmentation method
- image representation
- image quality