| AlexNet_ResNet_Concatenation Model Based on Rice Diseases Classification Enhancement |
| Paper ID : 1099-ICEEM2025 (R1) |
| Authors |
|
Medhat Hamdy Mansour *1, Heba Mohamed ElHoseny2, Adel Shaker El-Fishawy3, Sami Abdelmonem Eldolil4, Elhossiny Ibrahim Elhossiny5 1PH.D. student 2Faculty of Computer Studies, Arab Open University Riyadh 11681, Saudi Arabia 3dept. of Electronics and Electrical Communications Faculty of Electronic Engineering, Menoufia University Menouf 32952, Egypt 4dept. of Electronics and Electrical Communications Faculty of Electronic Engineering, Menoufia University Menouf 32952, Egypt 5dept. of Computer Science and Engineering Faculty of Electronic Engineering, Menoufia University Menouf 32952, Egypt |
| Abstract |
| Abstract— Rice is a vital global crop, serving as a staple food for over half of the world’s population. Early and accurate detection of rice infections is crucial to reducing yield losses and boosting production. While deep learning (DL)-based automated systems for disease diagnosis are advancing in precision agriculture, rice disease recognition remains a challenging research area. A key difficulty lies in the high visual similarity between symptoms of different diseases, making them hard to distinguish. Further research is needed to improve the differentiation of these complex and closely related diseases. DL techniques are considered the most relevant and appropriate techniques to address precisely and proficiently the difficulty of rice diseases recognition. Convolutional neural networks (CNNs) are the famous DL techniques used in rice disease classifications. Among such CNNs models, AlexNet and ResNet are considered the more suitable models to distinguish rice diseases. In this proposed research, a novel architecture is introduced to discriminate more precisely rice diseases based on concatenating the advantages of feature extraction capability in both AlexNet and ResNet models. The suggested architecture is evaluated on three public benchmarking datasets (DS) taken from famous websites including Mendeley and Kaggle which are DS1, DS2 and DS3, where the achieved results according to accuracy are 99.37 %, 99,86% and 98.27%, respectively, which eventually proves and asserts the superiority of our suggested work comparable to other state-of-the-art research studies. Keywords— DL, CNNs, feature extraction, AlexNet, ResNet. |
| Keywords |
| DL, CNNs, feature extraction, AlexNet, ResNet. |
| Status: Accepted |