Abstract
Wood chips are a leading renewable energy source and a vital raw material in pellet mills, bio-refineries, and paper mills. Their quality, especially for fuel and pulp applications, largely depends on accurately determining moisture content. With precise moisture content level measurements, manufacturers can fine-tune their processes to optimize quality and reduce waste, resulting in more efficient and sustainable operations. However, current data-driven methods used to measure the moisture content of wood chips need a large amount of data to train the models to eventually achieve state-of-the-art results. The major drawback is that collecting such a large dataset is time-consuming and the current approaches can take several days to do so. In this study, we propose a diffusion transformer, an advanced generative model that would be used to generate a variety of wood chip images with different moisture content levels. The proposed diffusion transformer demonstrates strong preliminary results, achieving high-quality generated images. The generated wood chip images serve as a crucial resource to train and enhance the robustness of computer vision models,