Due to the overpopulation and diversity of vehicle choices, Vietnam's urban traffic scene is extremely complex. Hence, it is challenging for autonomous vehicles to perform Object detection and Multiple object tracking (MOT) tasks. This paper tackles a part of the problem by focusing on the detection and tracking of vehicles. First, we propose a custom dataset of typical vehicles in Vietnam to improve detection accuracy. We also trained YOLOv8, a state-of-the-art detecting model, on the created dataset and achieved a mean Average Precision (mAP) of 0.87. For tracking, we implement tracking-by-detection paradigm methods, utilizing results from the trained YOLOv8 detector. By experimenting with recent robust tracking algorithms, StrongSORT and ByteTrack, in the context of Vietnam traffic, our results reveal the potential advantages of each tracker for the task. Our goal is to craft a custom pipeline that can perform efficient and reliable vehicle detection and tracking for autonomous driving in Vietnam.