Banca de QUALIFICAÇÃO: ANGELO LEITE MEDEIROS DE GOES

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : ANGELO LEITE MEDEIROS DE GOES
DATE: 01/11/2024
TIME: 09:00
LOCAL: Auditório do Núcleo de Pesquisa e Inovação em Tecnologia da Informação - nPITI
TITLE:

Framework for Implementing Intelligent Defect Inspection Systems in the Textile Industry

KEY WORDS:

Reference Architecture, Machine Learning, YOLO, Grid Detection, Ripple Refinement, TILDA Dataset.

 

PAGES: 118
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

This work presents the development of a reference architecture (RA) designed for automated quality inspection systems in the textile industry. The RA is structured into five layers: perception, middleware, service, application, and transversal quality attributes, encompassing all the physical and computational tools required for the implementation of these systems. A proof of concept was developed to validate the flexibility and feasibility of the proposed architecture. Two main algorithms were employed: grid detection, which segments images into patches for local defect classification, and ripple refinement, which improves detection quality by analyzing neighboring cells. Additionally, a benchmark of machine learning models was conducted, comparing eight models, including YOLOv5, YOLOv8 variants, as well as pre-trained networks LeNet5 and AlexNet, using the TILDA 400 dataset, composed of five distinct classes. The YOLOv8 models stood out, particularly YOLOv8 medium, which achieved an accuracy of 90.35% and an inference time of 0.5ms on a Tesla P100 GPU, significantly surpassing the 80% accuracy average obtained by human inspectors. The YOLOv8 small model also demonstrated a theoretical inspection potential of up to 46.875 m/min, far exceeding the maximum manual speed, which ranges between 15 to 20 m/min.

COMMITTEE MEMBERS:
Presidente - 347628 - ADRIAO DUARTE DORIA NETO
Externo ao Programa - 1241170 - HEITOR MEDEIROS FLORENCIO - UFRNExterno ao Programa - 2180207 - ITAMIR DE MORAIS BARROCA FILHO - UFRN