| An Improved Method for Apple Leaf Disease Detection Based on Hybrid DenseNet121 |
| Paper ID : 1077-ISCBAS |
| Authors |
|
asmaa mohamed * faculty of science, Al-Azhar university |
| Abstract |
| Plant diseases and pests significantly reduce the plant quality and yield. Hence, crop disease prevention and early detection are strategies that must be adopted in farming to save plants at an early stage and decrease overall food loss. Apples are the most lucrative fruit, but also the most susceptible to disease. Diseases affecting apple plants include rust and scab. Manual disease diagnosis can lead to incorrect pesticide identification and usage, which take a long time. Several deep learning algorithms have been used to address the challenge of identifying and classifying apple leaf diseases. However, these techniques are limited. This study proposes a hybrid model of DenseNet121 based on transfer learning and support vector machine (SVM) hinge loss. Its performance is then evaluated using the Plant Pathology 2020 dataset to identify the foliar disease categories in apple trees. To classify apple tree leaf diseases, the functionality of the pre-trained DenseNet121 model is used as the feature extractors, followed by the usage of an SVM classifier. Statistical analysis demonstrates that using a pre-trained DenseNet121 model and a supervised classifier algorithm to evaluate apple tree leaf diseases, especially apple scab, cedar apple rust, and multiple diseases, is quite advantageous. The proposed model outperforms previous state-of-the-art models in classification accuracy, precision, recall, and area under the curve values with 99% overall accuracy. |
| Keywords |
| agriculture, convolutional neural network, deep learning, support vector machine, transfer learning. |
| Status: Abstract Accepted (Oral Presentation) |