Name: | Description: | Size: | Format: | |
---|---|---|---|---|
7.53 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Com o crescimento exponencial dos telemóveis, esses dispositivos tornaram-se essenciais no cotidiano, auxiliando em diversas atividades e trazendo soluções práticas para
os desafios do dia a dia. Este trabalho explora o potencial destes equipamentos como ferramentas de apoio à segurança no trânsito, abordando especificamente o problema da
distração visual ao volante. Para isso, foi desenvolvido um sistema inovador que combina redes neurais convolucionais com a funcionalidade de dispositivos móveis. A metodologia
adotada focou na aquisição de um amplo conjunto de imagens para treinar um modelo de inteligência artificial capaz de classificar uma variável qualitativa em duas categorias
distintas: atenção e distração do motorista. Em particular, o estudo concentrou-se na criação de uma aplicação móvel que utiliza a câmera do telemóvel para monitorizar o
motorista e emitir alertas sonoros caso detecte distração prolongada. Os resultados obtidos destacaram a eficácia do modelo, especialmente após sua otimização para o formato
TensorFlow Lite, adequado para dispositivos móveis. Com uma execução 11,6 vezes mais rápida que o modelo padrão e uma redução de tamanho de 18,5 MB para 3,86 MB, evidenciando alta eficiência em velocidade e consumo de recursos.
With the exponential growth of smartphones, these devices have become essential in daily life, assisting in various tasks and providing practical solutions to everyday challenges. This paper explores the potential of smartphones as tools to support traffic safety, specifically addressing the issue of visual distraction while driving. To this end, an innovative system was developed that combines convolutional neural networks with the functionality of mobile devices. The adopted methodology focused on collecting a broad set of images to train an artificial intelligence model capable of classifying a qualitative variable into two distinct categories: driver attention and distraction. In particular, the study focused on creating a mobile application that uses the smartphone’s camera to monitor the driver and issue auditory alerts if it detects prolonged distraction. The results obtained highlighted the model’s effectiveness, especially after its optimization to the TensorFlow Lite format, making it suitable for mobile devices. With a runtime 11.6 times faster than the standard model and a size reduction from 18.5 MB to 3.86 MB, it demonstrates high efficiency in speed and resource consumption.
With the exponential growth of smartphones, these devices have become essential in daily life, assisting in various tasks and providing practical solutions to everyday challenges. This paper explores the potential of smartphones as tools to support traffic safety, specifically addressing the issue of visual distraction while driving. To this end, an innovative system was developed that combines convolutional neural networks with the functionality of mobile devices. The adopted methodology focused on collecting a broad set of images to train an artificial intelligence model capable of classifying a qualitative variable into two distinct categories: driver attention and distraction. In particular, the study focused on creating a mobile application that uses the smartphone’s camera to monitor the driver and issue auditory alerts if it detects prolonged distraction. The results obtained highlighted the model’s effectiveness, especially after its optimization to the TensorFlow Lite format, making it suitable for mobile devices. With a runtime 11.6 times faster than the standard model and a size reduction from 18.5 MB to 3.86 MB, it demonstrates high efficiency in speed and resource consumption.
Description
Mestrado de dupla diplomação com a Universidade Tecnológica Federal do Paraná
Keywords
Segurança no trânsito Redes neurais convolucionais Aplicativos móveis Detecção de distração