Identification of Corn Leaves Diseases Images Using MobileNet Architecture in SmartPhones

Abstract

Corn is most salient vegetable crop that contributes to the global economy. On the other hand, corn leaves diseases produce have a destructive impact into corn manufacture. Indeed, the farmers detect corn leaf diseases by observing principally the color change into corn leaves which is normally risky caused by the subjectivity and also it is spending time consumption. Over this scenario, it is necessary to develop computer vision algorithms that help to classify in an automatic and fast way corn leaves diseases in its early phase. Recent developments in the area of deep learning (DL) have considerably enhanced the accuracy of image classification. Thus, in this paper, a DL model approach is presented to detect and classify corn leaves diseases by taking images from a smartphone. The MobileNet architecture was chosen in our approach due to its versatility to be deployed in embedded systems. This architecture was adjusted in order to detect three kind of corn leaf diseases and when a corn leaf is healthy. The training and validation result observations have show the viability and effectiveness of the presented approach achieved more than 90% of accuracy.