Rice leaf disease poses a significant challenge to Jhum cultivation, making early and accurate detection vital for effective disease management. This study examines two cutting-edge deep learning models, VGG19 and DenseNet201, to identify and classify diseases in rice leaves. Our collection includes images of rice leaves that have been meticulously classed as bacterial blight, blast, brown spot, and the tungro. The diagnostic efficacy of each model was assessed after training with this dataset. Our findings reveal that the DenseNet201 model outperforms with a test accuracy of 99.63% and outstanding ROC-AUC scores across all disease categories. While the VGG19 model also demonstrates commendable performance with a test accuracy of 89.53%, it falls short of the DenseNet201 model. The findings highlight the potential of utilizing deep learning to transform disease detection in Jhum rice farming, providing valuable advantages for managing crops and optimizing yield.