Optimizing Leaf Disease Detection in Agriculture using Deep Learning

The agricultural industry is crucial for providing high-quality food and significantly contributes to the growth of both economies and populations. However, plant diseases can lead to substantial losses in food production and reduce species diversity. By utilizing precise and automated recognition methods, timely diagnosis of plant diseases can increase food production quality and decrease economic losses. Recently, deep learning has greatly advanced the accuracy of image classification and object detection systems, enhancing their effectiveness. Agriculture is the backbone of any country’s economy and is vital for the people’s survival. To ensure high crop production efficiency, it is essential to prevent plant diseases. For this study we considered a dataset with total of 15 classes which includes both healthy and diseases classes of leaves of total 6 categories i.e Apple healthy, Apple cedar rust, Apple scab, Potato early blight, Potato late blight, Tomato Early blight, Tomato Late blight, Grape healthy, Grape black rot, Corn (maize) rust, Corn (maize) healthy, Cherry healthy, Cherry powdery mildew. The Dense Convolutional Neural Network (DCNN) is employed for classification. In this paper, a pre-trained neural network model, DenseNet-121 is applied on the dataset. The dataset is trained by freezing layers in the DenseNet architecture to increase the time complexity. The proposal aims to optimize resource information to improve outcomes without added complexity. This results a model with 99.2% accuracy and 0.3% loss. Fine-tuning layers were added to the DenseNet-121 model for achievement of enhanced and higher results.