Describing Urban Events Using Deep Learning Techniques
Tales Lima Fonseca1; Lucas Arantes Berg1; Leonardo Goliatt da Fonseca1; Marcos de Mendonça Passini1
1 Universidade Federal de Juiz de Fora
doi:10.20906/CPS/CILAMCE2017-1269
Resumo
With the development of new information science technologies, large amounts of data are being generated every day in the urban environment, making it necessary to use computational techniques for its processing and analysis. In some cases the analysis depends on a human agent to interpret and evaluate a given situation. When devices generate data continuously, the amount of information that must be processed is limited by the performance of the worker and can result in a potential loss of quality in interpreting tasks. Considering this aspect, the use of intelligent devices can contribute considerably the improvement of the services of the traffic management and monitoring bodies. One example is intelligent traffic lights that identify vehicle flow through sensors and define how long they should be open in order to minimize local traffic jams or even to classify an image based on the object or in a certain situation. Artificial intelligence techniques have been shown to be efficient in this type of task, because they have the ability to analyze large datasets and make decisions in a short period of time. This article presents a tool that implements the concepts of Smart Cities and Artificial Intelligence in public data from the organs of traffic management of the city of Juiz de Fora. A deep learning neural network was developed to analyze data from monitoring cameras for traffic events.
Palavras-chave: Smart Cities; Deep Learning; Traffic; Cameras