A carregar...
Publicação
Implementation of Deep Learning models for Information Extraction on Identification Documents
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 9.64 MB | Adobe PDF |
Orientador(es)
Resumo(s)
The development of object detection models has revolutionized the analysis of personal information
on identification cards, leading to a decrease in external human labor. Although previous strategies
have been employed to address this issue without using machine learning models, they all present
certain limitations, which artificial intelligence aims to overcome. This report delves into the
development of a deep learning-based object detection capable of recognizing relevant information
from Portuguese identification cards. All the decisions made during the project will be accompanied
by a detailed background theory. Additionally, we provide an in-depth analysis of Optical Character
Recognition (OCR) technology, which was utilized throughout the project to generate text from
images. As the newest member of the Machine learning Team of Biometrid, I had the privilege of
being involved in this project that led to the improvement of the current approach that does not
leverage machine learning in the detection of relevant sections from ID cards. The findings of this
project provide a foundation for further research into the use of AI in identification card analysis.
Descrição
Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science
Palavras-chave
Artificial Intelligence Machine Learning Artificial Neural Networks Computer Vision Object Detection Models Optical Character Recognition (OCR)