[Webinar] A tail as old as “spine”: using 3D generative AI to explore squamate vertebrae past and present
(c) Katherine Wolcott (with contribution of Ed Stanley)
Vertebrae are abundant in the fossil record, yet rarely used to identify species. What if you could take a fragmented fossil vertebra, automatically reconstruct missing parts, and predict both its species and its exact position along the spine?
In this webinar, Katherine Wolcott will show how she and her collaborators did just that using DeepSDF, a generative AI model grounded in statistical shape modeling, trained on nearly 3,000 3D squamate vertebrae and achieving 96% accuracy in genus-level prediction and ±2 vertebra precision for spinal position. After briefly covering the theoretical foundations of micro-CT, deep learning, and vertebral morphology, there will be an interactive demo where participants can use the model to perform classification and shape completion on sample datasets. All materials will be provided; some Python experience is recommended, but not required.
Come explore how 3D generative AI can unlock new insights from ancient bones!
Speaker: Katherine Wolcott
Language: English