XAI Based Machine Learning approach for Bean Crop Disease Detection

Bean crops are vulnerable to numerous diseases such as anthracnose, bacterial blights, common bean mosaic, rust, etc. which can greatly affect their yield and quality. This research paper explores the diagnosis of bean crop diseases using machine learning algorithms by leveraging two feature extraction methods, namely Local Binary Pattern (LBP) and Discrete Wavelet Transform (DWT). These techniques were applied to extract significant features followed by simulating with various ML models for disease detection such as SVM, KNN etc. Additionally, Explainable AI (XAI) methodologies have been integrated into the dataset, providing transparency and interpretability to the performance of the machine learning models. Out of several ML models used, Random Forest performed the best with an accuracy of 0.81. Effect of various color spaces on the classification process have also been explored in this work and it was found that the model performed best in Greyscale color space. Here, the extracted features have been transformed into 1D feature vectors which makes the process computationally less expensive and as grayscale is a 2D color space, it adds more to the efficiency.