Uncertainty-based method for improving poorly labeled segmentation datasets.
Ekaterina RedekopAlexey ChernyavskiyPublished in: CoRR (2021)
Keyphrases
- segmentation method
- preprocessing
- segmentation algorithm
- high accuracy
- fully automatic
- energy function
- similarity measure
- test data
- prior knowledge
- significant improvement
- prior information
- synthetic data
- experimental results on real world
- data sets
- segmentation errors
- detection method
- cost function
- level set
- shape prior
- pixel level
- grey level
- thresholding method
- decision trees