Banana cultivation is an important source of economic income for several developing countries, with economies heavily dependent on their agricultural production.Such is the case of Colombia, a banana export country in which much of its production is still carried out by small peasant families without a high degree of modernization.The plant is quite sensitive to diseases caused by environmental conditions, bacteria, and viruses.Infections can spread rapidly and cause great damage to the plantations.For this reason, it is necessary to develop high performance and very low-cost technologies capable of quickly identifying the damage, to control it and reduce losses.In this article, we propose a convolutional model based on a deep neural network for the classification of banana plant leaf damage from images.The model is trained specifically for this problem with real images captured in different states of affectation of leaves of the plant.The model is suitable to be propagated on a very low-cost embedded system, and therefore suitable to be used in small plantations.The performance of the model demonstrates a high capacity to differentiate the damage to the leaves and the cause of it, making it possible to quickly formulate a treatment strategy.