GAN-Based Image Generation for Balancing Tomato Disease Image Datasets

Image classification algorithms are often trained on class-imbalanced datasets due to practical constraints in real-world data collection. This study uses Generative Adversarial Networks (GANs) to generate additional images of tomato plant organs to augment the training datasets to compensate for underrepresented classes, with the aim of improving the prediction accuracy of CNN classifiers. After analyzing the class distribution and the confusion matrix of a baseline classifier trained on the initial dataset, several GAN models are trained to produce realistic, class-specific tomato disease images that are added to the minority classes. The improvement from GAN-based augmentation is evaluated using standard classification metrics across multiple imbalance levels, yielding positive benefit in all cases between 0.8-1.4% depending on initial data availability. The presented augmentation pipeline, LeaFF-Balance, is intended to be generalizable to other tomato disease classification projects and adaptable to other plant species.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-14
DOI
https://doi.org/10.5281/zenodo.24219655
Primary Topic
Smart Agriculture and AI
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

GAN-Based Image Generation for Balancing Tomato Disease Image Datasets

Kerem Günay
Zenodo (CERN European Organization for Nuclear Research)
Smart Agriculture and AI
preprint

GAN-Based Image Generation for Balancing Tomato Disease Image Datasets

Kerem Günay
preprint en

Abstract

Image classification algorithms are often trained on class-imbalanced datasets due to practical constraints in real-world data collection. This study uses Generative Adversarial Networks (GANs) to generate additional images of tomato plant organs to augment the training datasets to compensate for underrepresented classes, with the aim of improving the prediction accuracy of CNN classifiers. After analyzing the class distribution and the confusion matrix of a baseline classifier trained on the initial dataset, several GAN models are trained to produce realistic, class-specific tomato disease images that are added to the minority classes. The improvement from GAN-based augmentation is evaluated using standard classification metrics across multiple imbalance levels, yielding positive benefit in all cases between 0.8-1.4% depending on initial data availability. The presented augmentation pipeline, LeaFF-Balance, is intended to be generalizable to other tomato disease classification projects and adaptable to other plant species.

Zenodo (CERN European Organization for Nuclear Research)
Yeditepe University (TR)
Smart Agriculture and AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon