Densenet 169-Based Plant Disease Detection of PlantVillage Dataset

Abstract

Abstract Using the PlantVillage dataset and a DenseNet-169-based spatial attention module, this article introducesa unique method for plant disease diagnosis. Because plant diseases can have a major influence onagricultural productivity, prompt intervention depends on precise identification.The intricacies and variances found in plant disease photos are frequently too much for conventionaltechniques to handle. We use DenseNet-169's strong feature extraction capabilities and supplement themwith a spatial attention module that highlights the most pertinent areas of the images in order to overcomethese difficulties. Additionally, we use fine-tuning methods to maximize our model's performance. Byfine-tuning, the DenseNet-169 architecture may better adjust to the unique features of the PlantVillagedataset, increasing its accuracy and resilience. When paired with spatial attention and fine-tuning,DenseNet-169 performs better than baseline models, attaining higher classification accuracy across arange of plant illnesses. Our results demonstrate how well DenseNet-169 may be integrated withfine-tuning and spatial attention processes for the diagnosis of plant diseases. This approach has thepotential to significantly improve crop management and yield by increasing detection accuracy andfostering more automated and dependable agricultural operations.

Plant phenotyping relevance

植物病害画像から病害状態を推定する深層学習手法の開発が中心であり、植物表現型（病害状態）の画像ベース推定に該当する。

Code and data availability

The paper uses the public PlantVillage dataset from Kaggle as its sole phenotyping image input for plant disease detection. No author code, trained models, or supplementary assets are reported.