Abstract

An ML-based system for interactive labeling of image datasets is contributedin TensorBoard Projector to speed up image annotation performed by humans. Thetool visualizes feature spaces and makes it directly editable by onlineintegration of applied labels, and it is a system for verifying and managingmachine learning data pertaining to labels. We propose realistic annotationemulation to evaluate the system design of interactive active learning, basedon our improved semi-supervised extension of t-SNE dimensionality reduction.Our active learning tool can significantly increase labeling efficiencycompared to uncertainty sampling, and we show that less than 100 labelingactions are typically sufficient for good classification on a variety ofspecialized image datasets. Our contribution is unique given that it needs toperform dimensionality reduction, feature space visualization and editing,interactive label propagation, low-complexity active learning, human perceptualmodeling, annotation emulation and unsupervised feature extraction forspecialized datasets in a production-quality implementation.

 

Introduction (beta)

None

 

Conclusion (beta)

None