MACHINE LEARNING FOR FACIAL RECOGNITION APPLIED TO LOCKS INTELLIGENT

• DOI: https://doi.org/10.22533/at.ed.3174272407115

• Palavras-chave: Computer vision. Deep learning. Convolutional neural networks. Eigenfaces. Fisherfaces.

• Keywords: Computer vision. Deep learning. Convolutional neural networks. Eigenfaces. Fisherfaces.

• Abstract:

  Security solutions for protecting property, such as conventional, smart and electronic locks, are available on the market. Some make use of biometric identification by means of fingerprints; however, facial recognition, using computer vision techniques through deep learning, has emerged as a promising alternative. The aim of this research is to analyze facial recognition algorithms applied to the problem of smart locks, using convolutional neural networks and evaluation metrics to measure the performance of the models generated. The nature of the research is applied and technology-based, using machine learning algorithms for facial recognition in smart locks. In terms of objectives, this is a descriptive study, as it evaluates the performance of the models generated by the methods and architectures employed using quality metrics in artificial neural networks. The results showed the promising effectiveness of the DenseNet-121 and DenseNet-169 architectures, combined with the Fisherfaces facial recognition method, in terms of precision, accuracy and F1, both in small-scale and large-scale analyses.