eduzhai > Applied Sciences > Engineering >

Connectivity-informed Drainage Network Generation using Deep Convolution Generative Adversarial Networks

  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: Stochastic network modeling is often limited by high computational costs togenerate a large number of networks enough for meaningful statisticalevaluation. In this study, Deep Convolutional Generative Adversarial Networks(DCGANs) were applied to quickly reproduce drainage networks from the alreadygenerated network samples without repetitive long modeling of the stochasticnetwork model, Gibb s model. In particular, we developed a novelconnectivity-informed method that converts the drainage network images to thedirectional information of flow on each node of the drainage network, and thentransform it into multiple binary layers where the connectivity constraintsbetween nodes in the drainage network are stored. DCGANs trained with threedifferent types of training samples were compared; 1) original drainage networkimages, 2) their corresponding directional information only, and 3) theconnectivity-informed directional information. Comparison of generated imagesdemonstrated that the novel connectivity-informed method outperformed the othertwo methods by training DCGANs more effectively and better reproducing accuratedrainage networks due to its compact representation of the network complexityand connectivity. This work highlights that DCGANs can be applicable for highcontrast images common in earth and material sciences where the network,fractures, and other high contrast features are important.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×