1Research Scholar, Department of Computer Science, Dr. N.G.P. Arts and Science College, Coimbatore, Tamilnadu, India
Email: shivashankariphd@gmail.com
ORCID: https://orcid.org/0009-0008-8106-5055
2Professor, Department of Computer Applications, Dr. N.G.P. Arts and Science College, Coimbatore, Tamilnadu, India
Email: kumuthini@drngpasc.ac.in
Foliar diseases have a considerable impact on the productivity of sesame (Sesamum indicum), an economically significant oilseed crop, leading to significant losses in both yield and quality. This paper suggests a computationally effective method for classifying sesame leaf diseases based on the ConvNeXt deep learning architecture in order to facilitate prompt and precise disease diagnosis for precision agriculture. To improve robustness in a variety of field settings, preprocessing and normalization are conducted to RGB leaf photos that have been enlarged to 224 × 224 × 3. While layer normalization stabilizes feature distributions, depthwise convolution extracts spatial features like lesions, discolouration, and abnormalities in texture. Residual connections maintain low-level information while guaranteeing stable gradient flow, and channel enlargement with GELU activation and subsequent channel reduction enhance discriminative feature learning. A fully linked layer and Softmax classification come after global average pooling, which combines spatial data into compact feature vectors. Backpropagation and cross-entropy loss are used to train the model. Experimental results demonstrate superior performance compared to conventional machine learning and earlier CNN-based techniques, achieving 97.72% overall accuracy, 97.85% True Negative Rate, 0.987 ROC-AUC, and 0.973 Cohen's κ. The suggested system offers a dependable, scalable, and real-time solution for automated sesame leaf disease diagnosis and sustainable crop management with a low inference latency of 12 ms and a moderate computing complexity of 4.3 GFLOPs.
Keywords: Leaf disease detection, Precision agriculture, ConvNeXt, Deep learning, Image classification, Feature extraction, Disease classification.
How to cite this article: Sivasankari K, Kumuthini C. Edge-Friendly Convnext Framework For Automated Sesame Leaf Disease Recognition. Int J Drug Deliv Technol. 2026;16(6s): 462-474; DOI: 10.25258/ijddt.16.6s.70
Source of support: None
Conflict of interest: None