At Syngenta, image classification is key to the modeling of samples within an automated chemical formulation system that has produced more than 100,000 pictures. The images are used to determine the next steps in the development of products but their interpretation by eye alone is very time-consuming and subjective to the evaluator.

The Torch Deep Learning Add-In for JMP Pro uses the power of convolutional neural networks and GPU computing to dramatically simplify the analysis and give consistent results across an agreed set of control images. We show how easy it is in the add-in to import and predict a rather complicated set of pictures and return meaningful, interactive output in minutes for tasks that normally take hours or even days with manual processing and Python coding.