Visión por computadora para detectar enfermedades fúngicas en cacao: revisión sistemática de arquitecturas de deep learning

This systematic literature review analyzes 43 scientific studies on the application of deep learning (DL) techniques for the early detection of diseases in cocoa (Theobroma cacao). Using the PRISMA methodology and the PICOC framework, models and architectures applied to classification, detection, an...

Descripción completa

Guardado en:
Autores principales: Yomer Ysidro-Santos, Andy Roel Vasquez-Bueno, Isait Camizán-Ojeda
Formato: Artículo
Lenguaje:inglés
Publicado: 2026
Materias:
Theobroma cacao
fungal plant disease
computer vision
deep learning architectures
YOLO
vision transformer
U-Net
disease detection
plant disease classification
transfer learning
explainable AI
self-supervised learning
lightweight models
smart agriculture
cacao cultivation
Smart Agriculture and AI
Advanced Neural Network Applications
Industrial Vision Systems and Defect Detection
Acceso en línea:Enlace permanente (DOI o Handle)
Descargar el archivo
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Sumario:This systematic literature review analyzes 43 scientific studies on the application of deep learning (DL) techniques for the early detection of diseases in cocoa (Theobroma cacao). Using the PRISMA methodology and the PICOC framework, models and architectures applied to classification, detection, and segmentation tasks were examined. The results show that lightweight models from the YOLO family achieve accuracy rates of up to 96% in cocoa disease identification, with the capability to perform real-time inference. Likewise, Vision Transformer (ViT) and U-Net architectures achieve accuracy rates ranging from 97% to 99% in controlled environments and related agricultural datasets, demonstrating their potential for transfer learning applications. However, limitations remain regarding cross-domain generalization, model explainability, and the availability of data collected under real-world field conditions. It is concluded that future research should expand public datasets, incorporate explainable artificial intelligence (XAI) and self-supervised learning (SSL), and optimize models for resource-constrained devices, thereby promoting sustainable and scalable solutions for smart cocoa cultivation.

Ejemplares similares