Orateur
Dr
Martin Jones
(The Crick Institute)
Description
The current generation of volume electron microscopy techniques routinely produce datasets in the terabyte regime. Modern computational techniques such as deep learning have shown a great deal of promise for automating the analysis of such data, but a major limitation is the lack of availability of ground truth training data which is often painstakingly generated manually by researchers. Our “Etch a Cell” projects on the Zooniverse have demonstrated that non-experts can produce data of sufficient quality for both direct downstream analysis and as training data for deep learning. This unprecedented quantity of high-quality training data raises tantalising prospects for genuine automation of in-depth analysis across a range of experiments.
