MedMNIST-C: Comprehensive benchmark and improved classifier robustness by simulating realistic image corruptions.
Francesco Di SalvoSebastian DoerrichChristian LedigPublished in: CoRR (2024)
Keyphrases
- single image
- image data
- input image
- image features
- template matching
- image retrieval
- image content
- real world
- image analysis
- image pixels
- low level
- region of interest
- classifier training
- image classification
- image representation
- image matching
- image collections
- intra class
- face detection
- berkeley segmentation dataset
- similarity measure
- image segmentation
- feature extractor
- morphological features
- labeled images
- sea surface
- spatial information
- pixel classification
- pixel values
- image set
- support vector machine
- multiscale
- training data
- learning algorithm