Identification and classification of leaf pests within the Indonesian mango farms using machine learning

Indonesia being the fourth largest producer of mangoes worldwide, agriculture is considered to be one of the major contributors to the Indonesian economy. However, in the past few years there has been a fall in the contribution of the agricultural field in the national Gross Domestic Product (GDP). The infestation of the Indonesian Mango farms by harmful pests have had a direct impact on the country’s economy. And the use of incorrect insecticides and pesticides can double the chances of having a poor yield from the crop and also indirectly affect animals and humans consuming the harvest. Machine Learning techniques applied to large agricultural datasets are capable of extracting valuable insights that can aid farmers in order to perform diagnosis of the leaf diseases. To achieve this, SVM and boosting based models like XGBoost and CatBoost are used for classification. Apart from that, a CNN built from scratch is built. Various experiments were conducted to evaluate the model performance. Convolutional Neural Network outperformed the other models with accuracy of 72.06%. This CNN model was taken as a baseline to create a python-based webpage to identify and classify infested mango leaf images