Evaluation of GAN Network-Based Images for Precision Agriculture

Precision agriculture has emerged as an alternative approach to optimize farming practices and process by giving advanced technologies base on data-driven decisions. In many cases, acquiring large labeled data sets for training deep learning models can be a time-consuming and expensive task. This paper discusses the evaluation of synthetic images generated by a trained GAN network using quantitative metrics. The proposed methodology can be divided into two parts; the design of the GAN networks, its training stage, and the generation of synthetic samples, and establishing the metrics for evaluating the quality of the synthetic images. The metrics calculated show that our synthetic generated images have an acceptable quality, slightly better when the Fruits360 data set is used.