Machine Learning System for the Evaluation of Tape Layup Scans

<p dir="ltr">The Convolutional Neuronal Network that is programmed in this Jupiter notebook can be trained to detect defects in surface scans of fibre tape layups. The network has a U-net architecture. The "transfer learning" method is used to first train the network with artificial data and then fine-tune it with real scans. The input data must have the format described in the benchmark data set (https://doi.org/10.25916/sut.27328356).</p><p dir="ltr">This publication contains the software as a Jupiter notebook.<br><br>It was created as part of the PhD project "Data-Driven Quality Assurance of the Dry Fibre Tape Laying Process." Funding: GA51557 under the Global Innovation Linkages program Round 2 by the Department of Industry, Science and Resources of the Australian Federal Government.</p>