Paul Zikopoulos, IBM | IBM Think 2019
summary
Demonstrates that a line-of-business user can build a 96% accurate video classifier without a data scientist by using Watson Machine Learning Accelerator with PowerAI Vision, clicking Auto Label and Augment Data to expand 119 frames to 4,249 training images. Shows NVLink-enabled Power Systems training a model in ~30 minutes versus 1.5 days on a laptop, and exposing the model as a REST API with one click; claims data-prep time drops from ~80% to ~20% and a developer adds a timer in half an hour.
AI-written summary · about
