Banca de QUALIFICAÇÃO: ANTONIO TEIXEIRA SANTANA NETO

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : ANTONIO TEIXEIRA SANTANA NETO
DATE: 31/03/2023
TIME: 08:00
LOCAL: online
TITLE:

 

CoffeeRust: a dataset for classifying the resistance of Coffea arabica to coffee leaf rust
 
 

KEY WORDS:

 

 
Stock Images, Coffea arabica, Genetic Improvement, Hemileia vastatrix, Convolutional Neural Networks, Computer Vision, Artificial Intelligence.
 

PAGES: 10
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUBÁREA: Eletrônica Industrial, Sistemas e Controles Eletrônicos
SPECIALTY: Automação Eletrônica de Processos Elétricos e Industriais
SUMMARY:

 

 
Coffee is the most important agricultural crop in the world and Brazil is one of its largest producers and consumers. The productivity of Coffea arabica crops is impacted by several factors, among the biotic ones is the coffee rust, caused by Hemileia vastatrix. Disease control is crucial and selection of resistant variants is an efficient strategy. In genetic improvement programs, resistance levels are identified using leaf discs inoculated with the pathogen and kept in a controlled environment. The samples are evaluated by an expert, which is subject to subjective analysis, in addition to being a time-consuming and costly process. Artificial intelligence (AI) seeks to develop algorithms that perform cognitive tasks similar to those performed by humans, simulating connections involving reasoning and learning. The convolutional neural network (CNN) is an AI technique frequently used for image processing and capable of extracting valuable information from the input data through computer vision techniques, allowing the realization of inferences with high precision and speed. However, the performance of these networks depends on the availability of a large set of training images. Although there are many databases on coffee rust, no large repositories were found that use the leaf disc methodology, nor with different stages of disease evolution in a standardized format. Therefore, this study proposes the development of a proprietary image bank that allows the use of learning transfer and CNNs to classify the degree of resistance of C. arabica cultivars to coffee rust. The proposed models showed accuracies greater than 95%, being more effective than traditional evaluation methods.
 

BANKING MEMBERS:
Presidente - 039.539.296-90 - LEONARDO BONATO FELIX - UFV
Externo ao Programa - 064.409.996-80 - HEVERTON AUGUSTO PEREIRA - UFV
Externa à Instituição - EVELINE TEIXEIRA CAIXETA MOURA
Notícia cadastrada em: 20/03/2023 09:25
SIGAA | NTInf - Núcleo de Tecnologia da Informação - | Copyright © 2006-2024 - UFSJ - sigaa05.ufsj.edu.br.sigaa05