Oral Presentation International Conference on Tomography of Materials & Structures

Application of Deep learning to the segmentation of synchrotron X‑ray tomography data of multiphase metal matrix composites (70)

Sergei Evsevleev 1 , Sidnei Paciornik 1 2 , Christian Gollwitzer 1 , Giovanni Bruno 1
  1. Bundesanstalt für Materialforschung und –prüfung (BAM), Berlin, BERLIN, Germany
  2. DEQM PUC-Rio, Rio de Janeiro, Brazil

The 3D microstructure of an Al alloy matrix composite with two ceramic reinforcements was investigated bysynchrotron X-ray tomography. Adeep learningalgorithm was usedforthesegmentation of four differentphases. We show that convolutional networkswith the U-Net architecture areable tosolve complex segmentation tasks with small amount of training data.

Download Full Paper