Development of a visual semantic descriptor applied to simultaneous localization and mapping

• Grant number: 21/08117-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: October 01, 2021
• End date: September 30, 2022
• Field of knowledge: Engineering - Electrical Engineering
• Principal Investigator: Valdir Grassi Junior
• Grantee: Gabriel Soares Gama
• Host Institution: Escola de Engenharia de São Carlos (EESC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 14/50851-0 - INCT 2014: National Institute of Science and Technology for Cooperative Autonomous Systems Applied in Security and Environment, AP.TEM

Abstract

The feature-based vSLAM (Visual Simultaneous Localization and Mapping) builds a map of an environment and locates itself simultaneously through the detection of features that stand out in it. The objective of this research is to develop a descriptor that uses the information inferred from a semantic segmentation neural network. This descriptor, despite having a higher computational cost, would have a greater amount of information stored and could increase the accuracy of the feature matching task, which would reduce the error present in the SLAM process. To obtain semantic information a neural network based on HRNet-OCR is going to be developed. The descriptor will be based on existing methods of extraction/description and semantic information will be added to complement the information already present. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

GAMA, GABRIEL SOARES; ROSA, NICOLAS DOS SANTOS; GRASSI JR, VALDIR; HOMEM, TPD; BIANCHI, RAD; DASILVA, BMF; CURVELO, CDF; PINTO, MF. Semantic SuperPoint: A Deep Semantic Descriptor. 2022 LATIN AMERICAN ROBOTICS SYMPOSIUM (LARS), 2022 BRAZILIAN SYMPOSIUM ON ROBOTICS (SBR), AND 2022 WORKSHOP ON ROBOTICS IN EDUCATION (WRE), v. N/A, p. 6-pg., 2022-01-01. (14/50851-0, 21/08117-0)