Bootstrapping Fine-Grained Classifiers: Active Learning with a Crowd in the Loop
We propose an iterative crowd-enabled active learning algorithm for building high-precision visual classifiers from unlabeled images. Our method employs domain experts to identify a small number of examples of a specific visual event. These expert-labeled examples seed a classifier, which is then iteratively trained by active querying of a non-expert crowd. These non-experts actively refine the classifiers at every iteration by answering simple binary questions about the classifiers’ detections. The advantage of this approach is that experts efficiently shepherd an unsophisticated crowd into training a classifier capable of fine-grained distinctions. This obviates the need to label an entire dataset to obtain high-precision classifiers. We find these classifiers are advantageous for creating a large vocabulary of visual attributes for specialized taxonomies. We demonstrate our crowd active learning pipeline by creating classifiers for attributes related to North American birds and fashion.
NIPS Workshop on Crowdsourcing: Theory, Algorithms and Applications, Lake Tahoe, 2013.