Active Learning
Uncertainty sampling
Least confidence: where the model selects those instances with the lowest confidence level to be labelled next.
Margin sampling: select instances where the margin between the two most likely labels is narrow, meaning that the classifier is struggling to differentiate between those two most likely classes.
Representative strategy
Information density in sample mean absolute difference or standard deviation between class size in crop
Papers
Geometry in active learning for binary and multi-class image segmentation
(1) maximum entropy of posterior probability distribution over classes, (2) minimal probability of selected class and (3) minimal gap between the two most probable classes.