Estimating crop yields with deep learning and remotely sensed data

This paper describes Illinois corn yield estimation using deep learning and another machine learning, SVR. Deep learning is a technique that has been attracting attention in recent years of machine learning, it is possible to implement using the Caffe. High accuracy estimation of crop yield is very important from the viewpoint of food security. However, since every country prepare data inhomogeneously, the implementation of the crop model in all regions is difficult. Deep learning is possible to extract important features for estimating the object from the input data, so it can be expected to reduce dependency of input data. The network model of two InnerProductLayer was the best algorithm in this study, achieving RMSE of 6.298 (standard value). This study highlights the advantages of deep learning for agricultural yield estimating.