Banca de Defesa de Proposta de Tese do Programa de Pós-graduação em Ciência da Computação. Confira abaixo as informações da atividade:
Aluno: ADRIEL SILVA DE ARAUJO
Título: INTEGRATING NORMAL VECTOR DATA WITH DEEP LEARNING FOR ENHANCED POINT CLOUD SEGMENTATION IN INTRAORAL SCANS
Orientador: Dr. Marcio Sarroglia Pinho
Banca Examinadora: Dr. Edgard Afonso Lamounier Junior (PPGEB/UFU), Dra. Isabel Harb Manssour (PPGCC/PUCRS)
Data: 08 de agosto de 2024
Local: Videoconferência
Horário: 15h30min
Resumo: The introduction of intraoral scanners (IOS) in orthodontics has transformed treatment planning and patient care by providing accurate 3D point cloud data of dental anatomy. This technology enhances diagnostic capabilities, improves communication among dental professionals, and streamlines linical procedures, leading to greater patient comfort and practice efficiency. Despite these advancements, accurate segmentation of anatomical structures within intraoral scans remains challenging, which is crucial for developing precise and personalized treatment plans. We propose to investigate the integration of normal vector data with deep learning models to advance point cloud segmentation in orthodontics. It focuses on improving the modified Difference of Normals approach by utilizing geometric information about surface orientations and curvatures. The research explores how to enhance segmentation accuracy by expanding handmade features and incorporating directional normal vector data into deep learning frameworks. The study also aims to develop novel processing modules and adapting existing neural network architectures to integrate normal vector data in this fashion, aiming to improve egmentation outcomes, bridging traditional techniques with modern deep learning approaches.
