Paper Details

Name: Satyaprakash Jena, Dhruv Gandhi, Amit Kumar Jaiswal & Harsh Khatri
Country: India

Year: 2026
Volume: Volume No: 13, January, Year: 2026 (Special Issue)
Page Number: 555-564
DOI: https://doi.org/10.5281/zenodo.19064623

ABSTRACT
Plants are at the center of where an economy is situated, the agric sector, and that of any nation's ecosystem. Look after your health to ensure it does not fall prey to many diseases caused by viruses, bacteria, and fungi. It entails due treatment, with detection of the same, and therefore must be conducted in such a manner as to put an end to irreplaceable damage to crops. Over the last few years, there has been spectacular progress in object discovery and image detection during deep network learning. Grounding on this, our research work aims to use pre-trained convolutional neural networks such as AlexNet, VGG16 and VGG19 via transfer learning for effective detection of plant disease. To facilitate improved model performance, we pre-process the images to improve the quality of the images and boost accuracy. Having trained the models, we tested them thoroughly to affirm the results. We are using the Plantvillage data set here, in which we have both a healthy leaf and a disease leaf. We split 80% of the training data and hold out 20% for the test. Along with accuracy, we also calculate accuracy, memory, and score F1 to check the models in general. The result confirmed that Alexnet achieved the outstanding test accuracy of more than 96.63%, which outperformed VGG16 (95.05%) and VGG19 (95.22%). Alexnet also achieved outstanding performance in other measurements at 92% precision, 91% memory and 91% F1 score. The above result confirms the efficiency of the AlexNet model trained to classify plant diseases with outstanding accuracy and efficiency. The aim of this work is the implementation of novel technology such as deep learning for crop protection to assist in achieving sustainable agriculture and economic development. Auto-detection of disease will help farmers act instantaneously to save crops, avoid loss and minimize the excessive use of chemical medication.

Index Terms – Deep Learning, Histogram equalization, RELU, Pooling, Fully Connected