Unsupervised Image Translation using Adversarial Networks for Improved Plant Disease Recognition

Nazki, Haseeb; Yoon, Sook; Fuentes, Alvaro; Park, Dong Sun

doi:10.48550/arxiv.1909.11915

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This can be explained on the basis of data quality, given that the synthetic data samples did not exactly replicate the original samples on a pixel-by-pixel basis, therefore requiring more instances in order to achieve superior performance. This research and consequent findings introduce new thinking in the applications of generative adversarial networks (Goodfellow et al, 2014) to plant disease classifier accuracy improvement since the current applications involve mostly data augmentation and to a less extent, resolving the problem of class imbalance (Nazki et al, 2019) while this endeavor directs the focus (with success) to the complete replacement of the entire original dataset with GAN-synthesized versions which can be theoretically supplied in an unlimited manner, with the only real limitation being storage capacity.…”

Section: Significance Of the Classification Resultsmentioning

confidence: 97%

Improvement of plant disease classification accuracy with generative model-synthesized training datasets

Albert

Bille²,

Leonard³

2023

RIB

View full text Add to dashboard Cite

Digitalization in agriculture requires critical research into applications of artificial intelligence to various specialization domains. This work aimed at investigating the application of image synthesis technology to the mitigation of the data volume constraint to digital plant disease phenotyping accuracy. We designed an experiment involving the use of a deep convolutional generative adversarial network (DC-GAN) to synthesize photorealistic data for healthy and bacterial spot disease-infected tomato leaves. The training dataset contained 1,272 instances per class. We further employed a 3-block visual geometry group (VGG) convolutional neural network (CNN) model with dropout regularization and 1 epoch to compare classification accuracies of the original dataset and various synthetic datasets. Our results showed that the third DC-GAN synthesized training dataset containing 3,816 synthetic examples of both healthy and bacterial spot infected tomato leaf classes outperformed the original training dataset containing 1,272 real examples of both tomato leaf classes (77.088% accuracy with the former dataset on a 3-block VGG CNN model with dropout regularization and 1 epoch, as compared to 76.447% accuracy with the latter dataset on the same classifier).

show abstract

Section: Significance Of the Classification Resultsmentioning

confidence: 97%