Detection and Classification of Diseases in Coffee Plant Using CNN-XGBoost Composite Model

Diseases in plants have always been a major concern for the agriculture community. Due to the absence of competent ways for detection, they suffer from loss of crop yield every year. This particular subset of ML, deep learning (DL) seemed to have shown much better performance in terms of accuracy. In this paper, we use the deep CNN architecture, InceptionNet and transfer learning for the detection and classification of plant diseases. Due to insufficient data availability, we have used DCGAN along with other traditional data augmentation techniques, to increase our dataset. This paper uses a InceptionNet-XGBoost composite model for classification. The said model is also compared against other deep CNN models, and seen to perform better. Moreover, the validity of XGBoost as the most suitable classifier is also shown. The composite model proposed has attained an accuracy of 98.32%. We have also precision, recall and F1-score as other performance metrics for the comparison of the model.