Efficient Detection of Tomato Leaf Conditions using Modified-InceptionResNetV2 Architecture

Pınar Uskaner Hepsağ

Article Details

Efficient Detection of Tomato Leaf Conditions using Modified-InceptionResNetV2 Architecture

Authors:

Pınar Uskaner Hepsağ

Publisher:

Osmaniye Korkut Ata Üniversitesi

Main Topics:

Derin Öğrenme | Nöral Ağlar

Journal:

Osmaniye Korkut Ata Üniversitesi Fen Bilimleri Enstitüsü Dergisi Vol.8 Issue 1, pp.375-392

Document Type:

Article

Publication Date:

1/2025

Document Language:

Turkish

Document ID (Org No.):

2001330021

pISSN:

2687-3729

Keywords:

Abstract:

 Timely identification and treatment of diseases affecting tomato leaves are essential for enhancing plant productivity, operational efficiency, and overall quality. Tomato plants are highly vulnerable to a diverse range of diseases, and farmers misdiagnosing of these ailments can lead to insufficient treatment strategies, causing harm to both the plants and the agroecosystem. Ensuring the quality of tomato crops relies significantly on prompt and accurate diagnoses. In contemporary times, deep learning techniques have demonstrated remarkable success across various applications, including classifying diseases in tomato plants. This study presents an approach for detecting tomato leaf conditions more precisely using a deep-learning architecture, namely the Modified-InceptionResNetV2 model, based on the InceptionResNetV2 transfer learning model. Our proposed architecture focuses on strengthening the classification block within the base model to achieve more accurate performance in identifying the condition of tomato leaves. Additionally, several preprocessing steps and augmentation techniques are employed to improve classification accuracy. Experimental analysis using a well-known, publicly available ten-class dataset achieves impressive training, validation, and testing accuracy rates of 99.74, 99.79, and 99.20, respectively. The proposed model could serve as a vital tool for farmers, aiding in the efficient detection and prevention of tomato diseases and enabling rapid and simple early detection of plant diseases. Experimental results showcase its superiority over previous studies in tomato leaf disease classification.

 Timely identification and treatment of diseases affecting tomato leaves are essential for enhancing plant productivity, operational efficiency, and overall quality. Tomato plants are highly vulnerable to a diverse range of diseases, and farmers misdiagnosing of these ailments can lead to insufficient treatment strategies, causing harm to both the plants and the agroecosystem. Ensuring the quality of tomato crops relies significantly on prompt and accurate diagnoses. In contemporary times, deep learning techniques have demonstrated remarkable success across various applications, including classifying diseases in tomato plants. This study presents an approach for detecting tomato leaf conditions more precisely using a deep-learning architecture, namely the Modified-InceptionResNetV2 model, based on the InceptionResNetV2 transfer learning model. Our proposed architecture focuses on strengthening the classification block within the base model to achieve more accurate performance in identifying the condition of tomato leaves. Additionally, several preprocessing steps and augmentation techniques are employed to improve classification accuracy. Experimental analysis using a well-known, publicly available ten-class dataset achieves impressive training, validation, and testing accuracy rates of 99.74, 99.79, and 99.20, respectively. The proposed model could serve as a vital tool for farmers, aiding in the efficient detection and prevention of tomato diseases and enabling rapid and simple early detection of plant diseases. Experimental results showcase its superiority over previous studies in tomato leaf disease classification.

 Timely identification and treatment of diseases affecting tomato leaves are essential for enhancing plant productivity, operational efficiency, and overall quality. Tomato plants are highly vulnerable to a diverse range of diseases, and farmers misdiagnosing of these ailments can lead to insufficient treatment strategies, causing harm to both the plants and the agroecosystem. Ensuring the quality of tomato crops relies significantly on prompt and accurate diagnoses. In contemporary times, deep learning techniques have demonstrated remarkable success across various applications, including classifying diseases in tomato plants. This study presents an approach for detecting tomato leaf conditions more precisely using a deep-learning architecture, namely the Modified-InceptionResNetV2 model, based on the InceptionResNetV2 transfer learning model. Our proposed architecture focuses on strengthening the classification block within the base model to achieve more accurate performance in identifying the condition of tomato leaves. Additionally, several preprocessing steps and augmentation techniques are employed to improve classification accuracy. Experimental analysis using a well-known, publicly available ten-class dataset achieves impressive training, validation, and testing accuracy rates of 99.74, 99.79, and 99.20, respectively. The proposed model could serve as a vital tool for farmers, aiding in the efficient detection and prevention of tomato diseases and enabling rapid and simple early detection of plant diseases. Experimental results showcase its superiority over previous studies in tomato leaf disease classification.

 Domates yapraklarını etkileyen hastalıkların zamanında tespit edilmesi ve tedavi edilmesi, bitki üretkenliğini, operasyonel verimliliği ve genel kaliteyi artırmak için esastır. Domates bitkileri çeşitli hastalıklara oldukça duyarlıdır ve çiftçilerin bu hastalıkları yanlış teşhis etmeleri, yetersiz tedavi stratejilerine yol açarak hem bitkilere hem de tarım ekosistemine zarar verebilir. Domates mahsullerinin kalitesinin sağlanması, zamanında ve doğru teşhise büyük ölçüde bağlıdır. Günümüzde derin öğrenme teknikleri, domates bitkilerinde hastalıkları sınıflandırmak gibi çeşitli uygulamalarda önemli başarılar göstermiştir. Bu çalışma, Modifiye-InceptionResNetV2 modeli adlı bir derin öğrenme mimarisi kullanarak domates yaprak koşullarını daha hassas bir şekilde tespit etmek için bir yaklaşım sunmaktadır; bu model, InceptionResNetV2 transfer öğrenme modeline dayanmaktadır. Önerilen mimari, temel model içindeki sınıflandırma bloğunu güçlendirmeye odaklanarak domates yapraklarının durumunu daha doğru bir şekilde tanımlama performansı elde etmeyi amaçlamaktadır. Ayrıca, sınıflandırma doğruluğunu artırmak için çeşitli ön işleme adımları ve artırma teknikleri kullanılmaktadır. Bilinen bir kamu veritabanı olan on sınıflı bir veri seti kullanılarak yapılan deneysel analiz, sırasıyla etkileyici eğitim, doğrulama ve test doğruluk oranlarına ulaşmaktadır: 99.74, 99.79 ve 99.20. Önerilen model, çiftçiler için önemli bir araç olarak hizmet edebilir; domates hastalıklarının etkili bir şekilde tespit edilmesine ve önlenmesine yardımcı olarak bitki hastalıklarının hızlı ve basit erken teşhisini sağlar. Deneysel sonuçlar, domates yaprak hastalığı sınıflandırmasında önceki çalışmalara üstünlüğünü ortaya koymaktadır.

Article Details

Efficient Detection of Tomato Leaf Conditions using Modified-InceptionResNetV2 Architecture

Authors:

Pınar Uskaner Hepsağ

Publisher:

Osmaniye Korkut Ata Üniversitesi

Main Topics:

Derin Öğrenme | Nöral Ağlar

Journal:

Osmaniye Korkut Ata Üniversitesi Fen Bilimleri Enstitüsü Dergisi Vol.8 Issue 1, pp.375-392

Document Type:

Article

Publication Date:

1/2025

Document Language:

Turkish

Document ID (Org No.):

2001330021

pISSN:

2687-3729

Tomato Leaf Diagnosis

Convolutional Neural Network(CNN)

Deep Learning (DL)

InceptionResNetV2

Modified-InceptionResNetV2Evrişimli Sinir Ağı (CNN)

Derin Öğrenme (DL)

Domates yaprak teşhisi

InceptionResNetV2

Modifiye-InceptionResNetV2

Issues

Article Details

Authors: Pınar Uskaner Hepsağ

Publisher: Osmaniye Korkut Ata Üniversitesi

Main Topics: Derin Öğrenme | Nöral Ağlar

Journal: Osmaniye Korkut Ata Üniversitesi Fen Bilimleri Enstitüsü Dergisi Vol.8 Issue 1, pp.375-392

Document Type: Article

Publication Date: 1/2025

Document Language: Turkish

Document ID (Org No.): 2001330021

pISSN: 2687-3729

Issues

Authors:

Pınar Uskaner Hepsağ

Publisher:

Osmaniye Korkut Ata Üniversitesi

Main Topics:

Derin Öğrenme | Nöral Ağlar

Journal:

Osmaniye Korkut Ata Üniversitesi Fen Bilimleri Enstitüsü Dergisi Vol.8 Issue 1, pp.375-392

Document Type:

Article

Publication Date:

1/2025

Document Language:

Turkish

Document ID (Org No.):

2001330021

pISSN:

2687-3729