Agriculture is the key driving force for food security and financial stability of a nation and specially in the rural economy where farmers face several challenges including lack of proper guidance, unpredictable procurement prices, uncertain yield prediction and inefficient crop selection. The present study presents a scalable AI based approach, which includes CNN and ML models for better coordination of soil classification and crop yield prediction. An RGB-image-based CNN model was developed to classify the four main soil types—Red, Black, Clay, and Alluvial — with high accuracy. In parallel, a ‘Best Crop Prediction’ module was built with multiple linear regression and was trained with more than a decade agricultural data consisting of type of soil, crop yield and regional environment. The models are packaged within a user-friendly web app where farmers can upload pictures of their soil and response with their geographic information to obtain crop recommendations based on these models. It is a friendly system for the general public (non-clients), so that precision agriculture becomes an accessible practice for everyone. By automating soil testing and crop selection, it’s designed to fill knowledge and resource voids, enabling farmers to make better decisions that optimize productivity and sustainability.
The model obtained a soil classification accuracy of 94% and contributed to increasing the accuracy of the crop recommendation by up to 28% against traditional methods. The real time prediction and user-friendly in the web application. Field trials verified the applicability of the system in different farming zones, and thus its potential for wide implementation. Next steps will include adding IoT sensors, and dynamic environmental data for further optimization.