TALISMAN: Targeted Active Learning for Object Detection with Rare Classes and Slices using Submodular Mutual Information.
Suraj KothawadeSaikat GhoshSumit ShekharYu XiangRishabh K. IyerPublished in: CoRR (2021)
Keyphrases
- rare classes
- mutual information
- active learning
- object detection
- rare class
- similarity measure
- feature selection
- image registration
- training set
- learning algorithm
- machine learning
- supervised learning
- emerging patterns
- random sampling
- training examples
- computer vision
- imbalanced datasets
- labeled data
- multi class
- semi supervised
- generalization error
- class imbalance
- imbalanced class distribution
- semi supervised learning
- unlabeled data
- cost sensitive
- learning process
- feature vectors
- test set
- data sets
- rare events