A Comprehensive Study of Public Datasets Used in Precision Agriculture to Classify Diseased and Healthy Crop Leaves

Computer vision technologies have gained substantial interest in precision agriculture due to their potential to automate tasks in the crop production cycle, from planting to harvesting. However, the limited availability of public image datasets hinders the rapid development and evaluation of deep learning algorithms for these tasks. Since 2015, several image datasets have been released to address this challenge. This paper provides the first comprehensive study of publicly available image datasets captured under field conditions for precision agriculture. It covers 12 datasets dedicatedly used in the field of leaf disease recognition only. The study highlights key dataset features and applications, emphasizing the need for high-quality datasets to support research in precision agriculture. This work offers valuable guidance on dataset selection and identifies areas where new datasets are needed to advance the field.