Leaf-Based Cacao Diseases Classification Using Image Processing - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue (Data Paper) Sci.Int.(Lahore) Année : 2024

Leaf-Based Cacao Diseases Classification Using Image Processing

Résumé

Cacao's diversified value as food and non-food ingredient increased its demand both in local and global markets; however, the presence of diseases affects its production. Identification of plant diseases is an important means of preventing losses in the yield and quantity of any agricultural crop, such as cacao. Manual means of identification and monitoring of diseases in cacao leaves is the common route, however, this requires a tremendous amount of work, expertise in plant diseases, and excessive processing time. Hence, an image processing technique in detecting cacao plant diseases through its leaves is presented in this paper. The cacao plant disease detection is conducted through the use of convolutional neural networks (CNN). The pre-trained CNN model, specifically the VGG19 model, is used as a feature extractor and classifier. Diseases under test are Swollen shoot, VSD (Vascular-Streak Dieback), and Witches Broom, in addition to healthy cacao leaves. Experimental results in this research show that the VGG19 model has achieved an accuracy of 88.75% for classifying cacao diseases through its leaves. The leaf-based classification via image processing of plant diseases in cacao could help farmers in early detection and intervention. However, an increased accuracy would be more viable if the number of images collected will be increased.
Fichier sous embargo
Fichier sous embargo
0 2 27
Année Mois Jours
Avant la publication
jeudi 25 juillet 2024
Fichier sous embargo
jeudi 25 juillet 2024
Connectez-vous pour demander l'accès au fichier

Dates et versions

hal-04483097 , version 1 (01-03-2024)

Identifiants

  • HAL Id : hal-04483097 , version 1

Citer

Dominic O Cagadas, Ruth April Labajan. Leaf-Based Cacao Diseases Classification Using Image Processing. Sci.Int.(Lahore), In press. ⟨hal-04483097⟩
17 Consultations
3 Téléchargements

Partager

Gmail Facebook X LinkedIn More