The research on detection of crop diseases ranking based on transfer learning

Abstract

Crop diseases are a major global threat to food security. Because the lack of agriculture experts or necessary facilities, it is difficult to determine the type of disease, as well as the degree of disease in time, which became the major factor affecting in crop production. In recent years, with the development of the transfer learning in deep learning domain, the experience of experts can be simulated to detect crop diseases in time. In this paper, we have proposed an improved transfer learning method based on ResNet 50 in crop disease diagnosis. The AI Challenger 2018 dataset has been deeper analyzed, the degree of crops diseases are detected. Comparing with non-transfer learning, the proposed transfer learning method achieved better results, which can significantly improve accuracy results by 5.1%~1.87% with reducing half of the running time. 2019 IEEE.