Design of an automatic image dataset generator for deep learning and its application to walnut species

Cihat Özil; Mursel Onder

Makale detayları

Design of an automatic image dataset generator for deep learning and its application to walnut species

Yazarlar:

Cihat Özil | Mursel Onder

Yayıncı:

Nigde Omer Halisdemir University

Ana Başlıklar:

Görüntü İşleme | Derin Öğrenme

Dergi:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.14 Issue 3, pp.899-906

Belge Tipi:

Makale

Yayın Tarihi:

7/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001354711

pISSN:

2564-6605

Anahtar kelimeler:

Özet:

 Traditional walnut grading methods have limitations such as dependence on expert knowledge, time-consuming processes and subjective evaluations. This situation increases the need for automatic and precise classification systems in agriculture and food industry. In order to fulfil this need, in this study, a unique and high resolution image dataset was created for the classification of five different walnut varieties Bilecik, Chandler, Fernette, Yavuz and Niksar widely grown in Turkey. For this purpose, an automatic data acquisition system that allows 360-degree imaging of walnuts was designed and developed. It was realised with the integration of an ESP32 microcontroller, stepper motor and a high-resolution smartphone camera. Each walnut was imaged from 144 different angles in 5-degree steps on the horizontal and vertical axes. The images were preprocessed with contour cropping and background removal. The resulting dataset was organised into training, validation and test sets. This study provides a strong foundation for future research on walnut classification.

 Traditional walnut grading methods have limitations such as dependence on expert knowledge, time-consuming processes and subjective evaluations. This situation increases the need for automatic and precise classification systems in agriculture and food industry. In order to fulfil this need, in this study, a unique and high resolution image dataset was created for the classification of five different walnut varieties Bilecik, Chandler, Fernette, Yavuz and Niksar widely grown in Turkey. For this purpose, an automatic data acquisition system that allows 360-degree imaging of walnuts was designed and developed. It was realised with the integration of an ESP32 microcontroller, stepper motor and a high-resolution smartphone camera. Each walnut was imaged from 144 different angles in 5-degree steps on the horizontal and vertical axes. The images were preprocessed with contour cropping and background removal. The resulting dataset was organised into training, validation and test sets. This study provides a strong foundation for future research on walnut classification.

 Traditional walnut grading methods have limitations such as dependence on expert knowledge, time-consuming processes and subjective evaluations. This situation increases the need for automatic and precise classification systems in agriculture and food industry. In order to fulfil this need, in this study, a unique and high resolution image dataset was created for the classification of five different walnut varieties Bilecik, Chandler, Fernette, Yavuz and Niksar widely grown in Turkey. For this purpose, an automatic data acquisition system that allows 360-degree imaging of walnuts was designed and developed. It was realised with the integration of an ESP32 microcontroller, stepper motor and a high-resolution smartphone camera. Each walnut was imaged from 144 different angles in 5-degree steps on the horizontal and vertical axes. The images were preprocessed with contour cropping and background removal. The resulting dataset was organised into training, validation and test sets. This study provides a strong foundation for future research on walnut classification.

 Geleneksel ceviz sınıflandırma yöntemleri, uzman bilgisine bağımlılık, zaman alıcı süreçler ve subjektif değerlendirmeler gibi sınırlamalar içermektedir. Bu durum, tarım ve gıda endüstrisinde otomatik ve hassas sınıflandırma sistemlerine duyulan ihtiyacı artırmaktadır. Bu ihtiyacı gidermek amacıyla bu çalışmada, Türkiyede yaygın olarak yetiştirilen beş farklı ceviz cinsinin Bilecik, Chandler, Fernette, Yavuz ve Niksar sınıflandırılması için özgün ve yüksek çözünürlüklü bir görüntü veri seti oluşturulmuştur. Bu amaç doğrultusunda, cevizlerin 360 derece görüntülenmesini sağlayan otomatik bir veri toplama sistemi tasarlanmış ve geliştirilmiştir. ESP32 mikrodenetleyici, step motor ve yüksek çözünürlüklü bir akıllı telefon kamerasının entegrasyonuyla gerçekleştirilmiştir. Her ceviz, yatay ve dikey eksenlerde 5 derecelik adımlarla 144 farklı açıdan görüntülenmiştir. Görüntüler, kontur kırpma ve arka plan temizleme ile ön işleme tabi tutulmuştur. Elde edilen veri seti eğitim, doğrulama ve test kümelerine ayrılarak düzenli şekilde organize edilmiştir. Bu çalışma, ceviz sınıflandırmasına yönelik gelecek araştırmalar için güçlü bir altyapı sunmaktadır.

Makale detayları

Design of an automatic image dataset generator for deep learning and its application to walnut species

Yazarlar:

Cihat Özil | Mursel Onder

Yayıncı:

Nigde Omer Halisdemir University

Ana Başlıklar:

Görüntü İşleme | Derin Öğrenme

Dergi:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.14 Issue 3, pp.899-906

Belge Tipi:

Makale

Yayın Tarihi:

7/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001354711

pISSN:

2564-6605

walnut classification

walnut dataset

automatic data acquisition

360 degree imaging

ESP32 microcontrollerceviz sınıflandırması

ceviz veri seti

otomatik veri toplama

360 derece görüntüleme

ESP32 mikrodenetleyici

Sayı

Makale detayları

Yazarlar: Cihat Özil | Mursel Onder

Yayıncı: Nigde Omer Halisdemir University

Ana Başlıklar: Görüntü İşleme | Derin Öğrenme

Dergi: Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.14 Issue 3, pp.899-906

Belge Tipi: Makale

Yayın Tarihi: 7/2025

Belge Dili: Türkçe

Document ID (Org No.): 2001354711

pISSN: 2564-6605

Sayı

Yazarlar:

Cihat Özil | Mursel Onder

Yayıncı:

Nigde Omer Halisdemir University

Ana Başlıklar:

Görüntü İşleme | Derin Öğrenme

Dergi:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.14 Issue 3, pp.899-906

Belge Tipi:

Makale

Yayın Tarihi:

7/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001354711

pISSN:

2564-6605