Tuesday, January 23, 2018
12:00 pm


"What's New? Interpretable Discovery in Large Image Data Sets"
Kiri Wagstaff, Principal Researcher in Artificial Intelligence and Machine Learning, JPL MLS Group, JPL

Automated detection of new, interesting, unusual, or anomalous items in large image data sets could help detect new near-Earth asteroids, fresh impact craters on Mars, and other key phenomena that might otherwise be lost within a large archive.  Many image data analysis systems are turning to convolutional neural networks (CNN) to represent image content due to their success in achieving high classification accuracy rates.  However, CNN representations are notoriously difficult for humans to interpret.  In this talk, I will discuss a strategy that combines novelty detection with CNN image features and yields interpretable explanations of novel image content.  I will show examples of discovery in Mars rover and Earth ecology image data sets.

Contact Diane Goodfellow diane@cms.caltech.edu
Add this event to my calendar