eduzhai > Applied Sciences > Engineering >

3DMaterialGAN Learning 3D Shape Representation from Latent Space for Materials Science Applications

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: In the field of computer vision, unsupervised learning for 2D objectgeneration has advanced rapidly in the past few years. However, 3D objectgeneration has not garnered the same attention or success as its predecessor.To facilitate novel progress at the intersection of computer vision andmaterials science, we propose a 3DMaterialGAN network that is capable ofrecognizing and synthesizing individual grains whose morphology conforms to agiven 3D polycrystalline material microstructure. This Generative AdversarialNetwork (GAN) architecture yields complex 3D objects from probabilistic latentspace vectors with no additional information from 2D rendered images. We showthat this method performs comparably or better than state-of-the-art onbenchmark annotated 3D datasets, while also being able to distinguish andgenerate objects that are not easily annotated, such as grain morphologies. Thevalue of our algorithm is demonstrated with analysis on experimental real-worlddata, namely generating 3D grain structures found in a commercially relevantwrought titanium alloy, which were validated through statistical shapecomparison. This framework lays the foundation for the recognition andsynthesis of polycrystalline material microstructures, which are used inadditive manufacturing, aerospace, and structural design applications.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×