Review of the State of the Art of Transfer Learning for Plant Leaf Diseases Detection

Abstract

Plant leaf diseases can have a significantly negative influence on the quantity and quality of agricultural cultivation, as well as the safety of food production. Plant leaf diseases could potentially entirely prevent the harvest of grains in some situations. Therefore, it is extremely important from a pragmatic standpoint to look for quick, automatic, cheap, and accurate ways to detect plant leaf diseases. One of the well-known plant leaf disease detection approaches is deep learning. Deep learning has several drawbacks as a result of the huge amount of data required to train the network. When a dataset has inadequate photographs, performance falls. An approach called "Transfer Learning" is an extensively used method for addressing the shortcomings of a small dataset, the length of the training process, and improving the performance of the model. In this study, we investigated transfer learning for deep CNNs to improve the learning capability to recognize leaf disease. This survey focuses on categorizing and analyzing the recent developments in transfer learning for Deep CNN situations to enhance learning performance by reducing the need for extensive training data collecting.