A Deep Learning Approach for Classifying Dianthus Seeds
30th IEEE Jubilee International Conference on Intelligent Engineering Systems, INES 2026, Budapest, Macaristan, 2 - 04 Temmuz 2026, ss.201-206, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/ines69513.2026.11661266
- Basıldığı Şehir: Budapest
- Basıldığı Ülke: Macaristan
- Sayfa Sayıları: ss.201-206
- Ankara Üniversitesi Adresli: Evet
Özet
Since ancient times, Dianthus has been popular in horticulture and floriculture as an ornamental flower. Besides its aesthetic and commercial value, it is of medical importance due to its proven pharmacological properties. Dianthus, a genus of the Caryophyllaceae family, has species with different characteristics such as color, size, shape and fragrance. For plant research and seed industry, classifying seed species is a time-consuming, expensive, and crucial operation that requires specialist staff and equipment. This study used a variety of pre-trained convolutional neural networks (CNNs) to propose a deep transfer learning-based method for classifying Dianthus seed species. Model training and testing were conducted using a recently compiled dataset that included images of three Dianthus species: Dianthus caryophyllus, Dianthus chinensis, and Dianthus petraeus. With an AUC of 0.9996 and an accuracy of 99.45%, the ResNet152V2 architecture performed better than the others. Additionally, the model successfully depicts the morphological characteristics of every dianthus species, as demonstrated by Grad-CAM visualizations, which supports the models' dependability and explainability. For improved seed quality control, the suggested approach can be easily incorporated into botanical and agricultural applications.