In the heart of Turkey, at Selcuk University, a team of researchers led by Hakan Terzioğlu from the Department of Electrical-Electronics Engineering has been making waves in the agritech world. Their recent study, published in the journal *Agronomy* (translated to English as “Field Management”), is shedding new light on how artificial intelligence can revolutionize tomato disease detection, a critical challenge in global agriculture.
Tomatoes, one of the most widely consumed crops worldwide, are highly susceptible to a variety of pathogens. These diseases can lead to significant yield losses, impacting both farmers’ livelihoods and the global food supply. Traditional methods of disease detection often rely on manual inspection, which can be time-consuming and prone to human error. This is where Terzioğlu’s research comes into play.
The team constructed an original dataset comprising 6414 images captured under real production conditions. These images were categorized into three types: leaves, green tomatoes, and red tomatoes, encompassing five classes of diseases: healthy samples, late blight, early blight, gray mold, and bacterial cancer.
“We wanted to create a robust dataset that reflects real-world conditions,” Terzioğlu explained. “This is crucial for developing models that can perform accurately in practical agricultural settings.”
The researchers evaluated twenty-one deep learning models, selecting the top five performers for feature extraction. From each model, 1000 deep features were extracted, and feature selection was conducted using three different methods: MRMR, Chi-Square, and ReliefF. The top 100 features from each selection technique were then used for reclassification with traditional machine learning classifiers under five-fold cross-validation.
The results were impressive. The highest test accuracy of 92.0% was achieved with EfficientNet-b0 features, Chi-Square selection, and the Fine KNN classifier. “This combination of deep learning-based feature extraction with traditional classifiers and feature selection techniques proved to be highly effective,” Terzioğlu noted.
The study also found that EfficientNet-b0 consistently outperformed other models, while the combination of NasNet-Large and Wide Neural Network yielded the lowest performance. These findings highlight the potential of combining advanced deep learning techniques with traditional machine learning methods to enhance disease detection in agriculture.
The implications of this research are significant for the agricultural sector. Accurate and efficient disease detection can lead to timely interventions, reducing crop losses and improving yields. This, in turn, can enhance food security and support farmers’ economic stability.
Moreover, the integration of AI in agriculture aligns with the growing trend of smart farming, where technology is leveraged to optimize agricultural practices. As Terzioğlu’s research demonstrates, the future of agriculture lies in the synergy between cutting-edge technology and traditional methods.
“This research opens up new possibilities for the application of AI in agriculture,” Terzioğlu said. “We are excited to see how these findings can be translated into practical solutions that benefit farmers and the broader agricultural industry.”
As the world grapples with the challenges of climate change and food security, innovations like those from Terzioğlu’s team are more important than ever. By harnessing the power of AI, we can pave the way for a more sustainable and productive future in agriculture.