Indian Journal of Science and Technology
Year: 2022, Volume: 16, Issue: 27, Pages: 2047-2057
Bisrat Getnet1*, Desalegn Ashebir2, Birhanu Gardie1
1Department of Information Technology, School of Computing & Informatics, Mizan-Tepi University, Ethiopia
2Department of Software Engineering, School of Computing & Informatics, Mizan-Tepi University, Ethiopia
Email: [email protected]
Received Date:23 May 2022, Accepted Date:14 November 2022, Published Date:18 July 2022
Objective: Design a convolutional neural network model for flower disease detection that can recognize and categorize flower diseases according to their disease types. Method: A 10-way SoftMax classification system is used to group the majority of diseases that affect flowers into distinct disease classes (normal, powdery mildew, rose aphid, botrytis blight, downey mildew, red spider mites, Japanese beetles, rose resettle, gray mold, and black spot). A total of 4200 images—70% for training, 15% for validation, and 15% for testing—were used to train and validate the model’s performance. From the Tana Flora flower cultivation center in Bahir Dar Amhara, Ethiopia, we obtained a dataset. The proposed model was trained by generating additional images using image augmentation techniques to overcome over fitting problem. Findings: Our flower disease identification model produced cutting-edge results with a test accuracy of 94.67%. When used on the same dataset, state-of-the-art models like Alex Net’s test accuracy was 85.6%, Google Net’s test accuracy was 90.98, and VGG19’s test accuracy was 89.3%. Our flower disease detection model has been shown to be faster to train and has a smaller model size thanks to the median filter that we used to improve the quality of the photos and a novel segmentation method that fit our dataset. Novelty: We proposed a new CNN architecture and new segmentation algorithm to identify flower disease. Through our experiments, we have shown median filtering, and ways of segmentation improve the performance of our model.
Keywords: Deep learning; Feature Learning; Segmentation; Flower Disease; Median Filtering
© 2023 Getnet et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Published By Indian Society for Education and Environment (iSee)
Subscribe now for latest articles and news.