Test-time augmentation-based active learning and self-training for label-efficient segmentation.
Bella Specktor-FadidaAnna LevchakovDana SchonbergerLiat Ben-SiraDafna Ben-BashatLeo JoskowiczPublished in: CoRR (2023)
Keyphrases
- active learning
- uncertainty sampling
- cost sensitive
- experimental design
- semi supervised
- image labeling
- object segmentation
- medical imaging
- segmentation algorithm
- semi supervised learning
- segmentation accuracy
- medical images
- image analysis
- training set
- image segmentation
- label fusion
- co training
- machine learning
- active learning strategies
- random sampling
- fully automatic
- segmentation method
- level set
- relevance feedback
- multiscale
- image sequences
- decision trees