Iterative Active Classification of Large Image Collection

To efficiently and accurately classify a large image collection, this paper proposes a novel interactive system by incorporating active learning, online learning and user intervention. Given an image collection, our system iteratively alternates the interactive annotation and verification until all the images are classified. The main advantage is that it provides faster interactive classification rates than alternative approaches. Our system achieves this goal by a unified active learning algorithm that selects the images to be annotated or verified, which requires a probability model for simulating the time cost of human input during manual intervention. To assist manual annotation and verification, we generate the classification hypothesis of the selected images using a conditional random field (CRF) framework, which combines the cues from an online learned classifier and pairwise similarities of unlabeled images. Experimental results demonstrated the effectiveness of the method.