Ag-AI: A Dynamic AI Engine for Agriculture

Artificial Intelligence (AI) technology has been adopted to solve several agricultural problems including weed detection, plant disease identification, and crop quality prediction. However, not all AI models (machine learning (ML) and deep learning (DL)) exhibit the same performance for diverse agricultural datasets. Moreover, the model performance varies among the tasks (e.g., regression, and classification). However, selecting a robust and accurate AI model is a tedious task as it requires repeating the entire pipeline of data pre-processing, feature selection, model training, and hyperparameter optimization. In this paper, we propose a dynamic AI engine that automates the selection and development of AI models for agricultural use cases. The core of our AI engine is a reinforcement learning-based robust recommendation system that learns the optimal model by interacting with a given dataset. By eliminating the need to manually select the models, the proposed engine accelerates the adoption of AI by farmers, thereby increasing operational efficiency and enabling farmers to make informed decisions.