Deep Learning-Build Detection and Classification of Tomato Leaf Diseases Using Advanced CNN Architectures

Tomato leaf diseases pose a crucial threat to agricultural productivity, making the development of effective and accurate detection methods crucial. To address these challenges, our project utilizes deep learning for automated disease detection. By leveraging advanced feature extraction techniques, we eliminate the require for manual feature engineering and segmentation. A custom Convolutional Neural Network (CNN) developed for multi-class classification. The model involve of four convolutional layers with increasing filter sizes (32, 64, 128, and 256), utilizing ReLU and SELU activations for capable feature extraction. Max-pooling layers are employed to bring down the spatial dimensions of the data, followed by a fully connected layer with 1024 units. The final softmax layer classifies inputs into three distinct categories. The network is improved using the Adam optimizer with a learning rate of 0.0001 and trained with sparse categorical cross-entropy loss. The proposed model attained an accuracy of 94%, evidencing its effectiveness in classifying image data while maintaining a stabilty between computational efficiency and carrying out, offering a highly exact method for identifying tomato leaf infection.