PLANT DISEASE DETECTION USING DEEP LEARNING

HAFRI ,MOHAMMED ZIYAD; KASRI ,SOUMAYA

PLANT DISEASE DETECTION USING DEEP LEARNING

Date

2023

Authors

HAFRI ,MOHAMMED ZIYAD

KASRI ,SOUMAYA

Publisher

Faculty of Sciences

Abstract

This project aimed to develop two deep learning models Densnet121 and CNN for plant disease detection using a dataset of leaf images (tomato - pepper - potato). The primary objective was to achieve high accuracy in classifying and identifying diseases affecting at (tomato - pepper - potato). The dataset utilized was PlantVillage dataset is available on Kaggel, which consisted of 20,600 images (tomato - pepper - potato). The project involved several stages, including loading the dataset into a TensorFlow dataset and preprocessing the images using custom layers for resizing, rescaling, and data augmentation. The CNN model architecture comprised convolutional and pooling layers, followed by a flattening layer, dense layers, and a final classification layer. The model was compiled with the Adam optimizer, and accuracy was selected as the evaluation metric. To ensure reliable evaluation, the dataset was divided into training, testing, and validation sets. The normal cnn model was trained for 20 epochs using the training dataset, with a batch size of 32 and the Densenet121 was trained for 100 epochs The validation set was employed to assess the two models performance during training, and the training history was recorded for further analysis. The two trained model demonstrated an impressive accuracy achievement of 0.96 for the normal cnn model and for the Densenet121 0.98, highlighting its effectivenes the Densenet121 in accurately detecting and classifying plant diseases. Random images were used for testing, and the model consistently exhibited high precision in disease prediction better than the normal cnn model.