Deep learning in agriculture: A survey

This paper surveys the emerging application of deep learning, a modern technique for image processing and data analysis, within the domain of agriculture. The authors review 40 research efforts that employ deep learning to address a variety of agricultural and food production challenges, examining for each study the specific agricultural problem under investigation, the particular models and frameworks used, and the sources, nature, and preprocessing of the data, together with the overall performance achieved according to the metrics reported in each work.

The survey further compares deep learning against other existing and popular techniques with respect to differences in classification or regression performance. Its findings indicate that deep learning provides high accuracy and outperforms commonly used image processing techniques, supporting the view that deep learning is a promising, high-potential approach for agricultural problems and offering a consolidated map of the field that helps orient future research.