Scholarlink

This work aims to study a circuit design for the development of a digital pupillometer using an ESP board, and to explore tools for analyzing pupillary dynamics using deep learning methods, such as the U-net architecture for convolutional neural networks, applied for pupil segmentation in the obtained eye images. As for the neural network implementation, we developed a pre-trained U-net-based image segmentation model for pupillary analysis, which may be applied to the newly obtained data from the electronic device in development. We further intend to enhance the convolutional network in order to achieve improved metric scores and to better adapt the model for our specific purposes.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Project of a Low-Cost Digital Pupillometer Ran by Software Tools for Data Analysis Based on Deep Learning for Image Segmentation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Denillo G. C. de Barros,
  • Guilherme R. Unfried,
  • Maria F. O. de Figueiredo,
  • Matheus Y. S. Cruz,
  • Sara P. Borba,
  • Petros A. Samways,
  • Sergio Okida,
  • Cristhiane Gonçalves

摘要

This work aims to study a circuit design for the development of a digital pupillometer using an ESP board, and to explore tools for analyzing pupillary dynamics using deep learning methods, such as the U-net architecture for convolutional neural networks, applied for pupil segmentation in the obtained eye images. As for the neural network implementation, we developed a pre-trained U-net-based image segmentation model for pupillary analysis, which may be applied to the newly obtained data from the electronic device in development. We further intend to enhance the convolutional network in order to achieve improved metric scores and to better adapt the model for our specific purposes.