Multi-stage Transfer Learning for Corn Leaf Disease Classification

Abstract

Corn is one of the prime commodities in many parts of the world. However, corn yield is affected by natural environment factors such as weather, soil condition, humidity, and diseases. Using machine learning, this study proposes a multi-stage transfer learning for corn leaf disease classification and presents initial experiment results. Results show that InceptionV3 achieves 99% accuracy while Xception attains 96% accuracy, and InceptionResNetV2 performs at 94% accuracy. Also, the multi-stage transfer learning model is compared with other models considering quality measures such as accuracy and training time. This study indicates that the multi-stage transfer learning models developed is comparable with existing deep learning models. Future extension of this work is proposed to improve the performance of the corn leaf disease classification models.