Tomato Leaf Disease Detection and Classification using GAN and ResNet 9

Background: Tomato leaf diseases are a serious menace in agricultural production and require the best methods for accurate detection. Among deep learning techniques, Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN) seem to offer significant opportunities in disease diagnosis. Objective: This study evaluates the efficiency of ResNet-9 and DenseNet-121 architectures of Convolutional Neural Networks combined with synthetic data augmentation by Generative Adversarial Networks. Methods: ResNet-9 and DenseNet-121 frameworks assisted feature extraction and classification by adding pooling layers for downsizing dimensions and both ReLU activation and dropout layers for non-linear operation and overfitting prevention. Using GAN-based image synthesis produced additional synthetic data, which boosted the model performance. Results: The experimental results indicate that ResNet-9 performed better than DenseNet-121 by obtaining 97.02% accuracy with a validation loss of 0.0808. Conclusion: ResNet-9 demonstrates effective usage as a deep learning-based method to detect exact diseases in tomato leaves when implemented efficiently. Real-time disease control solutions through deep learning techniques will enable precision agriculture to enhance crop health thereby improving yield production.