Landmark detection and anthropomorphic measurements for calculation of biometric parameters used in ophthalmic lens fitting

• Grant number: 18/01189-3
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Duration: July 01, 2019 - June 30, 2020
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Marcos Makoto Ikegame
• Grantee: Marcos Makoto Ikegame
• Host Company: Optarium Comércio de Artigos de Ótica Eireli - ME
• CNAE: Comércio varejista de artigos de óptica ; Desenvolvimento e licenciamento de programas de computador não-customizáveis
• City: Campinas
• Associated researchers: Douglas Aparecido dos Santos
• Associated grant(s): 19/27836-8 - Development of Mobile App for calculation of biometric parameters for ophthalmic lens fitting, AP.PIPE
• Associated scholarship(s): 19/24015-3 - Landmarks detection and anthropomorphic measures to calculate biometric parameters used in the assembly of ophthalmic lenses, BP.TT; 19/12936-7 - Landmarks detection and anthropomorphic measures to calculate biometric parameters used in the assembly of ophthalmic lenses, BP.TT; 19/13811-3 - Landmarks detection and anthropomorphic measures to calculate biometric parameters used in the assembly of ophthalmic lenses, BP.TT; 19/12939-6 - Landmarks detection and anthropomorphic measures to calculate biometric parameters used in the assembly of ophthalmic lenses, BP.TT

Abstract

Precise measurements in order to properly fit ophthalmic lenses in eyewear frames is of vital importance to anyone who need to wear glasses. The use of computer systems for anthropometry in this field is limited to high cost apparatus that demands fixed references for properly calculating measurements. Furthermore, such systems need an human operator who manually chancels users' facial landmarks. Alternatively, most of the population has their lens fitting measurements taken by hand, through the use of tools developed in the early 20th century. The goal of this project is to develop a model allows for the detection of fiduciary points in users' faces and reckons - without the usage of fixed references of any kind - both main measurement demanded for properly fitting lenses in frames: naso-pupillary distance and fitting height. In order to achieve this goal, we shall adopt MTL (Multi-Task Learning) models; more precisely, we will rely on Deep Learning techniques through convolutional networks. We hope to develop a model from which we can detect landmarks and measure absolute anthropometric distances. Such system will allow for the diffusion of computer assisted measurement of biometric parameters for eyewear fitting and leaving imprecise human measurements in the past. (AU)