Are Labels Always Necessary for Classifier Accuracy Evaluation?
Weijian DengLiang ZhengPublished in: CVPR (2021)
Keyphrases
- training data
- class labels
- unseen data
- high accuracy
- fold cross validation
- training set
- confusion matrix
- highest accuracy
- training examples
- labeling effort
- individual classifiers
- classification rate
- roc curve
- improve the recognition accuracy
- label noise
- computational cost
- prediction accuracy
- error rate
- training samples
- significantly improves the accuracy
- semantic labels
- labeled instances
- learning algorithm
- false positives
- annotation effort
- decision trees
- active learning
- fully supervised
- support vector machine
- higher classification accuracy
- supervised classifiers
- supervised learning
- graph cuts
- class distribution