This article aims to investigate the leaf spot disease in rice, a condition that has been observed to cause substantial reductions in crop yield, ranging from 10% to 50%. The disease is characterized by the appearance of circular lesions on the leaves of rice plants, which gradually expand and cover the entire leaf surface, leading to the eventual death of the affected leaves. In order to achieve automation in the detection and severity classification of the disease under investigation, a range of deep learning models were utilized. These models encompassed Faster DenseNet121 and UNet. These methods proposed in this study was trained and tested using a dataset consisting of 1120 images. The findings of the study indicate that the DenseNet121 approach that yielded the highest level of success demonstrated a noteworthy classification accuracy rate of 95.23%. The results obtained from the integration of image processing techniques with deep learning models demonstrated superior performance compared to the analysis solely based on deep learning models. The integration described in this study presents a novel approach to enhance the efficiency of diagnosing leaf spot disease on rice. By leveraging advanced technology, this integration aims to reduce the time required for diagnosis, minimize the occurrence of human errors, and facilitate the timely application of suitable pesticides to mitigate the detrimental effects of the disease. © 2024 Elsevier B.V., All rights reserved.