RESUMO
The constant increase in the number of vehicles around the world has
drawn attention to the need to implement effective traffic monitoring
and control technologies. In this work, a methodology is presented
that uses two algorithms, one that applies the YOLOv8 detector and
the other that applies Deep SORT tracking, to detect, track and count
vehicles. Validation results during training of the vehicle detection
model revealed an overall accuracy of 89, 2% and mAP50 of 93, 7%
on the AIC HCMC 2020 database. Additionally, when applying the
model to a test video, an overall accuracy of 81, 2% and mAP50 of
87, 5% was achieved. These results highlight the effectiveness of the
detection algorithm for use in applications involving vehicle tracking
and counting, with significant potential for traffic management.