Abstract:
Transportation problems happen in cities because they are so inherent in diversity and always
changing. The result is that traffic congestion, accidents, and the economic burden of these
things are extremely difficult to deal with in the modern day. The economy is slowed down
due to these impediments, and the safety of individuals is put in jeopardy. As cities continue
to grow and concerns about traffic and congestion continue to rise, detecting vehicles and
humans can provide a potential route to a more effective solution. YOLOv8n, YOLOv8s,
YOLOv8m, and YOLOv8l are the four variants of YOLOv8 that have been utilized in this
study to investigate how well they function in detecting vehicles and persons. It has
been accomplished by collecting a fresh dataset of Dhaka's streets. The purpose of this
research is to address the problem mentioned above. The collection of data includes images
of humans as well as six different types of vehicles. By gaining a high mAP of 0.909 at IoU
50, the results demonstrate that YOLOv8l is successful in detecting vehicles and humans in
demanding traffic settings. Additionally, the results address the effectiveness of the dataset.
Based on this accuracy, it appears that the model has the potential to be effectively utilized
for a wide range of purposes, including the automation of traffic enforcement, the
improvement of traffic flow, the reduction of congestion, the prevention of accidents, and the
preparation of the groundwork for autonomous vehicles in Bangladesh.
