SegmentMeIfYouCan: A Benchmark for Anomaly Segmentation.
Robin ChanKrzysztof LisSvenja UhlemeyerHermann BlumSina HonariRoland SiegwartMathieu SalzmannPascal FuaMatthias RottmannPublished in: CoRR (2021)
Keyphrases
- berkeley segmentation dataset
- segmentation algorithm
- image segmentation
- quantitative evaluation
- medical images
- level set
- region growing
- segmentation accuracy
- fully automatic
- image analysis
- neural network
- word segmentation
- boundary detection
- segmented images
- brain mri
- segmentation method
- optimal segmentation
- intrusion detection
- joint segmentation
- grey level
- ground truth
- video segmentation
- multiscale
- anomaly detection
- multiple objects
- motion segmentation
- prior information
- test images
- fully unsupervised
- image regions
- energy function
- feature vectors
- benchmark suite