MCDAL: Maximum Classifier Discrepancy for Active Learning
This addresses the problem of efficient data labeling in active learning for computer vision researchers, offering a more stable alternative to GAN-based methods.
The paper tackles the instability and hyper-parameter sensitivity of GAN-based active learning methods by proposing MCDAL, a framework that uses prediction discrepancies between multiple classifiers to measure uncertainty for sample acquisition, achieving state-of-the-art performance on image classification and semantic segmentation datasets.
Recent state-of-the-art active learning methods have mostly leveraged Generative Adversarial Networks (GAN) for sample acquisition; however, GAN is usually known to suffer from instability and sensitivity to hyper-parameters. In contrast to these methods, we propose in this paper a novel active learning framework that we call Maximum Classifier Discrepancy for Active Learning (MCDAL) which takes the prediction discrepancies between multiple classifiers. In particular, we utilize two auxiliary classification layers that learn tighter decision boundaries by maximizing the discrepancies among them. Intuitively, the discrepancies in the auxiliary classification layers' predictions indicate the uncertainty in the prediction. In this regard, we propose a novel method to leverage the classifier discrepancies for the acquisition function for active learning. We also provide an interpretation of our idea in relation to existing GAN based active learning methods and domain adaptation frameworks. Moreover, we empirically demonstrate the utility of our approach where the performance of our approach exceeds the state-of-the-art methods on several image classification and semantic segmentation datasets in active learning setups.