Achieving 10,000x training data reduction with high-fidelity labels

This post was originally published by Source Author on Google.

Human-Computer Interaction and Visualization