Abstract: Many major grain-producing nations have implemented steps to limit their grain exports as COVID-19 has expanded globally; food security has sparked significant worry from several stakeholders. One of the most crucial concerns facing all nations is how to increase grain output. Crop infections, however, are a challenging issue for many farmers, thus it's critical to understand the severity of crop diseases promptly and properly to support staff in taking additional intervention steps to reduce plants being further affected. This paper proposed a hybrid deep learning model that combines the benefits of deep residual networks and dense networks to identify tomato leaf disease. This model can reduce the number of training process parameters to increase calculation accuracy and improve the gradients and information flow. Since the original RDN model was developed for image super-resolution, we must modify the input image characteristics and hyperparameters to reorganize the network architecture for classification tasks. On the Tomato test dataset in the AI Challenger 2018 datasets, experimental findings demonstrate that this model can obtain a top-1 average identification accuracy of 95%, confirming its good performance. In terms of crop leaf recognition, the reconstructed residual dense network model can outperform the majority of the state-of-the-art models while using significantly less computing power.
Index Terms: Agricultural artificial intelligence; tomato leaf diseases; residual dense network; identification of leaf diseases
|
DOI:
10.17148/IJARCCE.2023.124181
