A deep learning model for early prediction and diagnosis of late blight disease in irish potatoes.

Abstract

Introduction: According to the Centre International potato, the Irish potato is a popular crop in Uganda with a great value for income generation and improving nutrition. The Ugandan government has declared the Irish potato a key crop for the country. Despite Irish potatoes being a key crop in Uganda, diseases are a challenge in their growth, growth most especially the late blight disease. Problem: Late blight of Irish potato disease causes 60 percent damage to Irish potatoes and this has affected the production of Irish potatoes in Uganda due to misdiagnosis and inaccurate prediction of the disease. AI models have been providing significant results ill the improvement of disease prediction and diagnosis but Datasets of an African Origin and Uganda in particular still remain a challenge. The main objective: T110 main objective of this research is to develop a deep learning model powered mini pathogen Laboratory to improve early prediction and accurate diagnosis of late blight disease in Irish potatoes in Kigezi-Uganda. Methodology: The methodology used in this study was Design Science methodology and Artificial Intelligence approaches for model development. This achieved development of a deep learning model powered Irish potato Mini Pathogen lab. Results: This application achieved a lot of success in its evaluation with minimum accuracy of 94% and maximum accuracy ranging from 98% to 100% thus contributing to bridging the accuracy gap of East Africa which 6.5% to 7% by closing it by 5.5% and closing in to the global accuracy of 99.56% On particular experiments. Conclusion: The deep learning model powered Irish potato Mini Pathogen lab helped extension workers to conveniently and cheaply carryout experiments on late blight dis- ease early prediction and accurate diagnosis. Future works: Future research should be done in further crop science to experiment with the model and give specific recommended on specific herbicides to treat late blight. late and specific predict ion on spraying schedules. More investigation should be done in unleashing the black-box to make future models more explainable.