Virtual sawing using generative adversarial networks

Abstract

The global trend towards digitalization along with the 4th Industrial Revolution allows building new innovative solutions to optimize manufacturing. In particular, the highly competitive sawmill industry is not an exception. The industry always depends on efficient raw material utilization, and thus, exploration of the internal log structure is an important feature in the timber conversion process. One common approach for a comprehensive internal wood structure examination is virtual sawing that is predicting the outcome of sawing process based on log measurements. The main objective of this thesis was to study the suitability of state-of-the-art generative adversarial networks for virtual sawing. The specific aims were to choose an appropriate generative adversarial network architecture to build a trainable model as an extension to an existing virtual sawing system. The proposed method for image-to-image translation is capable to synthesize the photorealistic images of the boards based on log measurements. The defects (knots) on virtual boards were detectable and their locations correspond to those in real boards.