Active Learning for Structured Prediction from Partially Labeled Data
This work addresses the high labeling cost problem for researchers and practitioners in computer vision and structured prediction, though it is incremental as it builds on existing active learning frameworks.
The paper tackles the problem of reducing labeled data requirements for structured prediction tasks by proposing a novel active learning algorithm that selects examples to maximize expected information gain. Experiments demonstrate that the algorithm outperforms previous active learning methods and achieves accuracy comparable to fully supervised methods with significantly less labeled data, such as in human action and group activity recognition in videos.
We propose a general purpose active learning algorithm for structured prediction, gathering labeled data for training a model that outputs a set of related labels for an image or video. Active learning starts with a limited initial training set, then iterates querying a user for labels on unlabeled data and retraining the model. We propose a novel algorithm for selecting data for labeling, choosing examples to maximize expected information gain based on belief propagation inference. This is a general purpose method and can be applied to a variety of tasks or models. As a specific example we demonstrate this framework for learning to recognize human actions and group activities in video sequences. Experiments show that our proposed algorithm outperforms previous active learning methods and can achieve accuracy comparable to fully supervised methods while utilizing significantly less labeled data.