GeoMotor: Design with nature. Recognition of geometries using a convolutional neural-network approach (CNN)

Abstract

This article conceptualizes a solid regular called GeoMotor capable of moving and directing the sediments of a mountain river and changing its geography. The GeoMotor manages to manipulate the directional growth of sediments in an artificial environment, unveiling emerging architectural structures. For this, an analog simulation of the mountain river flow was performed and provide data to understand the phenomenon. Subsequently, this data was used to train a neural network that recognizes the emerging architectural patterns. As future work, it is planned to improve the models to offer functionalities beyond the orthodox practices of traditional architectonic models