Proposta de implementação de análise de vídeo com séries temporais
This work proposes an architecture implementation capable of classifying video segments in embedded security systems in a timely manner. Two possible architectures were evaluated, verifying their viability with the system, evaluating the training and validation accuracy rate, confusion matrix and...
Na minha lista:
Autor principal: | |
---|---|
Outros Autores: | |
Formato: | bachelorThesis |
Idioma: | pt_BR |
Publicado em: |
Universidade Federal do Rio Grande do Norte
|
Assuntos: | |
Endereço do item: | https://repositorio.ufrn.br/handle/123456789/48320 |
Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
Resumo: | This work proposes an architecture implementation capable of classifying video segments in embedded security systems in a timely manner. Two possible architectures were
evaluated, verifying their viability with the system, evaluating the training and validation
accuracy rate, confusion matrix and prediction execution time. A dataset was also created
for training the neural network, for classifying suspicious and non-suspicious activities.
Thus, it was possible to conclude that the architecture chosen for the final application had
a satisfactory performance, reaching about 90% of accuracy in validation data and, execution time, embedded in Raspberry Pi 4, average of 25 seconds, qualifying the architecture
for this possible application |
---|