Towards Sustainable Agriculture: A Stepwise Model Using Hybrid CNN Algorithms for Disease Detection

Abstract

Despite several obstacles including plant diseases, pollinator rise, and climate change, agriculture is vital to the economic growth of many nations and supports the worldwide population. Crop leaf diseases have grown to be a major worry for the agricultural sector, yet a lack of automated crop disease detection techniques makes prompt diagnosis and recognition difficult in many parts of the world. Plant diseases will increase food insecurity and have an impact on the nation's income if they are not recognized in a timely manner. The identification and control of crop diseases is essential for raising agricultural production, cutting expenses, and advancing ecologically friendly crop treatment practices. Crop disease detection systems that are automated have been developed using modern technology like machine learning algorithms and data mining. In order to build this system, three iterations of an improved Hybrid CNN algorithm were used, together with photos of crop pairings that were infected and healthy to create a stepwise disease detection model. The suggested model has been enhanced to maximize accuracy and speed of detection, and it has been used for plant leaf disease detection as well as multi-class crop identification. Two well-known publicly available dataset namely “PlantVillage '' and “CropImages” are used for experiment.  The obtained findings showed higher accuracy (96.0%), precision (95.0%), F1-scores (96.0%), and recall (94.0%) when compared to state-of-the-art methods. The article offers a cutting-edge analysis, details our approach and experimental findings, and offers suggestions and future possibilities for our study on the use of hybrid CNN models to the categorization of crop and plant leaf diseases.