PotatoGANs - Utilizing Generative Adversarial Networks, Instance Segmentation, and Explainable AI for Enhanced Potato Disease Identification and Classification

Sep 21, 2026

Uploaded by Jiven

Numerous applications enable agricultural disease segmentation via deep learning, yet performance often drops on new conditions due to overfitting. In potato farming, fast and accurate disease identification is essential for protecting yields and supporting agricultural economies. The research introduces PotatoGANs, a synthetic data augmentation framework that uses two GAN variants to generate disease images from healthy inputs, improving diversity and generalization. Experiments assess image realism with Inception Score, with CycleGAN exceeding Pix2Pix quality. The approach further enhances explainability by combining GradCAM, GradCAM++, and ScoreCAM with DenseNet169, ResNet152 V2, and InceptionResNet V2, and uses Detectron2 on the extended dataset for instance segmentation, achieving strong Dice performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

AccessFree to download

Abstract

Keywords

1 Introduction

arXiv :2405 .07332v2 [ cs .CV] 22 Jun 2025 PotatoGANs: Utilizing Generative Adversarial Networks, Instance Segmentation, and Explainable AI for Enhanced Potato Disease Identification and Classification Fatema Tuj Johora Faria 1†* , Mukaffi Bin Moin 1†, Mohammad Shafiul Alam 1†, Ahmed Al Wase 1 , Md. Rabius Sani 1 , Khan Md Hasib2 1 Ahsanullah University of Science and Technology, Dhaka, Bangladesh. 2 The University of Western Australia, Crawley, WA, Australia. *Corresponding author(s). E-mail(s): [fatema.faria142@gmail.com](fatema.faria142@gmail.com) Contributing authors: [mukaffi28@gmail.com](mukaffi28@gmail.com) ; [ahmed.alwasi34@gmail.com](ahmed.alwasi34@gmail.com) ; [rshridoy010113@gmail.com](rshridoy010113@gmail.com) ; [shafiul.cse@aust.edu](shafiul.cse@aust.edu) ; [khanmdhasib.aust@gmail.com](khanmdhasib.aust@gmail.com) ; †These authors contributed equally to this work. Abstract Numerous applications have resulted from the automation of agricultural disease segmentation using deep learning techniques. However, when applied to new conditions, these applications frequently face the difficulty of overfitting, resulting in lower segmentation performance. In the context of potato farming, where diseases have a large influence on yields, it is critical for the agricultural economy to quickly and properly identify these diseases. Traditional data augmentation approaches, such as rotation, flip, and translation, have limitations and frequently fail to provide strong generalization results. To address these issues, our research employs a novel approach termed as PotatoGANs. In this novel data augmentation approach, two types of Generative Adversarial Networks (GANs) are utilized to generate synthetic potato disease images from healthy potato images. This approach not only expands the dataset but also adds variety, which helps to enhance model generalization. Using the Inception score as a measure, our experiments show the better quality and realisticness of the images created by PotatoGANs, emphasizing their capacity to resemble real disease images closely. The CycleGAN model outperforms the Pix2Pix GAN model in terms of image quality, as evidenced by its higher IS scores CycleGAN achieves higher Inception scores (IS) of 1.2001 and 1.0900 for black scurf and common scab, respectively. This synthetic data can significantly improve the training of large neural networks. It also reduces data collection costs while enhancing data diversity and generalization capabilities. Our work improves interpretability by combining three gradient-based Explainable AI algorithms (GradCAM, GradCAM++, and ScoreCAM) with three distinct CNN architectures (DenseNet169, Resnet152 V2, InceptionResNet V2) for potato disease classification. This comprehensive technique improves interpretability with insightful visualizations and provides detailed insights into the network’s decision-making. The goal of combining several CNN designs and explanation techniques is to maximize interpretability while offering an in-depth understanding of the model’s behavior. We further employ this extended dataset in conjunction with Detectron2 to segment two classes of potato disease images, with the primary goal of enhancing the overall performance of our model. Furthermore, in the ResNeXt-101 backbone, Detectron2 has a maximum dice score of 0 .8112. This combination of PotatoGANs and Detectron2 has the potential to be a powerful approach for tackling the limitations of traditional data augmentation approaches while also boosting the accuracy and robustness of our disease segmentation model. Keywords Generative Adversarial Networks, CycleGAN, Pix2Pix GAN, Instance Segmentation, Smart Farming, Convolutional Neural Network, Potato Disease Detection, Detectron2, Explainable Artificial Intelligence 1 Introduction Agriculture is the most important factor in ensuring global food security. It serves as the foundation of sustenance, producing the supp

FAQs

What problem does PotatoGANs address in agricultural disease segmentation?

It targets the reduced segmentation performance caused by overfitting when deep learning models face new conditions beyond their training data.

How does PotatoGANs generate training data for potato disease recognition?

It uses two types of Generative Adversarial Networks to generate synthetic potato disease images from healthy potato images, expanding the dataset and increasing variety.

How is interpretability improved in the proposed potato disease classification pipeline?

Three gradient-based Explainable AI methods—GradCAM, GradCAM++, and ScoreCAM—are combined with three CNN architectures (DenseNet169, ResNet152 V2, and InceptionResNet V2) to provide visual explanations of the model’s decisions.

Sep 21, 2026

Uploaded by Jiven