Diagnosing Floral Diseases Automatically using Deep Convolutional Neural Nets

In floriculture, quick and precise disease detection is critical to crop health and profitability. Traditional procedures are frequently labour-intensive and inefficient, which highlights the need for new methods. This study solves this issue by using improvements in AI and image processing via Deep Convolutional Neural Networks (DCNNs). The goal was to create a strong deep learning model capable of identifying common flower illnesses using a dataset of $\mathbf{1, 6 0 2}$ high-resolution photos spanning five disease classes: ‘healthy,’ ‘rhizopus,’ ‘powdery mildew,’ ‘rust,’ and ‘blight’. The contribution of this work is the development of an ensemble model that leverages the strengths of the Xception and DenseNet121 architectures to improve classification accuracy. Transfer learning to adapt pretrained models using ImageNet weights for this specific task, thereby enhancing the model’s generalization capabilities. The ensemble learning strategy increased classification accuracy by $\mathbf{3 . 7 \%}$, resulting in a final accuracy of $96.4 \%$, which is significantly higher than utilizing independent models. This model demonstrates the effectiveness of integrating deep learning, transfer learning, and data augmentation, even with minimal datasets. The findings show that this approach has the potential to revolutionize disease detection in floriculture, allowing for more efficient crop health monitoring, higher yields, and increased economic viability. The use of these advanced techniques demonstrates Artificial intelligent rising involvement in agriculture, with this model helping to develop more effective and scalable disease detection systems.