An improved YOLO11 model for accurate cotton growth-stage detection

Accurate crop detection in orchard environments is one of the primary challenges for agricultural systems today, as vision plays a crucial role in modern solutions for plant monitoring, crop harvesting, early disease detection, and analysing water and nutritional status, among other applications. In...

ver descrição completa

Detalhes bibliográficos
Autores: González Valdivia, Guillem, Foix Salmerón, Sergi|||0000-0001-9249-6696, Alenyà Ribas, Guillem|||0000-0002-6018-154X
Formato: capítulo de livro
Fecha de publicación:2025
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:dnet:upcommonspor::1eaa78772949c66acf6da9304a78d9f7
Acesso em linha:https://hdl.handle.net/2117/460908
https://dx.doi.org/10.3233/FAIA250601
Access Level:acceso abierto
Palavra-chave:Deep learning
Custom YOLO
Cotton detection
Cotton dataset
Classificació INSPEC::Pattern recognition::Computer vision
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
Descrição
Resumo:Accurate crop detection in orchard environments is one of the primary challenges for agricultural systems today, as vision plays a crucial role in modern solutions for plant monitoring, crop harvesting, early disease detection, and analysing water and nutritional status, among other applications. In our case, we want to detect cotton bolls and accurately label them in different stages of growth. To address this challenge, we present a custom YOLO11 model with a selection of state-of-the-art vision modules to improve cotton detection and reduce possible mismatches with a custom cotton dataset. Our approach achieves better detection accuracy with around 7M parameters and 16 GFLOPS, less than the official YOLO versions’ small to medium variants (8, 10, 11, 12). The final model will be used to count and monitor cotton plants in a greenhouse.