Detection of Plant Disease From Plant Image


Creative Commons License

Kılıç R., Sağlam H. K., Oral E. A., Özbek İ. Y.

ISASE2018, Erzurum, Türkiye, 26 - 28 Kasım 2018 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Erzurum
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: Classification, Feature extraction, Image processing, Plant disease, Symptom
  • Atatürk Üniversitesi Adresli: Evet

Özet

Plants are of great importance in providing ecological balance. They are not only a source of nutrients but also a source of oxygen. Therefore, plants need to be protected. By looking at the appearance of a leaf one can decide what disease it has. With early detection of the disease in plants, excessive damage can be prevented. For this purpose we implemented a method to detect plant diseases from picture of its leaves. This method is based on deep learning classification which is carried out by convolutional neural networks. Various experiments were performed using the PlantVillage database. This database includes 54,305 images for 12 different plants for which there exists one healthy and different number of disease classes each. There is total of 38 classes with 26 different diseases. We obtained 98.47% success rate for the classification of diseases.