FieldPlant: A Dataset of Field Plant Images for Plant Disease Detection and Classification With Deep Learning

The Food and Agriculture Organization of the United Nations suggests increasing food supply by 70% to feed world population 2050, although approximately one third all is wasted because plant diseases or disorders. To achieve this goal, researchers have proposed many deep learning models help farmers detect in their crops as efficiently possible avoid yield declines. These are usually trained on personal public disease datasets such PlantVillage PlantDoc. composed laboratory images captured under conditions, with leaf each a uniform background. dataset very low accuracies when running field complex backgrounds multiple leaves per image. solve problem, PlantDoc was built using 2,569 downloaded from Internet annotated identify individual leaves. However, includes some absence pathologists during annotation process may resulted misclassification. In study, FieldPlant suggested that 5,170 collected directly plantations. Manual image performed supervision ensure quality. This 8,629 across 27 classes. We ran various benchmarks evaluate state-of-the-art classification object detection found tasks outperformed those