DefinePK

Abstract

The planting and commercialization of coffee is an important source of economic resources and commercial dynamizer for many developing countries, particularly with economies that are strongly dependent on agricultural production, as is the case of Colombia. Coffee is the most important export product of the country and enjoys a high reputation for its quality and flavor. Although the country has done a lot of research to develop the sector, investment in technology is very low, and most of its cultivation for export (of the highest quality) is done by small coffee families without a high degree of technology, and without major resources to access it. The quality of the coffee bean is strongly sensitive to various diseases induced by environmental conditions, fungi, bacteria, and insects, which directly and strongly affects the economic income of the entire production chain and the country. In many cases the diseases are rapidly transmitted, causing great economic losses. A quick and reliable diagnosis would have an immediate effect on reducing losses, which is why the development of a low-cost embedded system capable of making reliable diagnoses in the hands of peasant farmers is proposed. In this article, we propose the development of a software model capable of identifying in real-time the possible disease of a plant from an image of a leaf. For this, we use a DenseNet convolutional neural network trained with 1250 images corresponding to five categories that include the most important diseases of the plant. Laboratory tests show that the proposed model is capable of operating on a low-cost embedded system with a high-performance rate by correctly categorizing the plant's leaves against an unknown set.