Crop Yield Prediction via CNN-Transformer Architecture Leveraging Satellite Imagery and Weather Data

It is a challenging task of predicting crop yields because several factors influence it; hence accurate prediction is important to allow farmers to manage their crop production to meet the growing population food demands. This paper proposes a novel method for estimating agricultural produce based on meteorological data and satellite imagery in a CNN-Transformer model. In contrast, the use of temporal and geographical characteristics of satellite data as well as meteorological parameters allows the model to provide correct yield estimates. The proposed CNN-Transformer model received an accuracy of 89.5%, proving that it can solve information in agriculture, which tends to have multiple attributes. The actual data was used to evaluate the efficiency of developed model and its applicability for making decisions in agriculture. Moreover, the system is portable, and can be used for more crop types and geographical regions. This activity not only rewards sustainable agricultural practices but also enriches the existing guidelines for crop production probability forecasting based on the AI technique.