HETRA: A Real-Time Traffic Detection and Directional Counting System Using YOLOv8 and DeepSORT

Rapid urbanization and economic growth in developing countries such as India have increased the demand for automated traffic analysis systems. However, heterogeneous traffic conditions present significant challenges to automated traffic analysis. This study proposes HETRA (HEterogeneous TRaffic Analysis), a real- time traffic monitoring framework that integrates YOLOv8 and DeepSORT for multi-class detection, tracking, and counting. For this, a custom traffic dataset was developed using real-world recorded traffic videos. The dataset consists of more than eight thousand annotated frames representing the fourteen traffic classes (motorized, non-motorized, and pedestrians). The proposed model leverages YOLOv8 for class identification and localization, while DeepSORT is used for multi-class tracking and directional traffic counting. The custom traffic dataset was used to train and assess the proposed HETRA model. The model achieved precision, recall, and mean average precision values of 71.15%, 65.51%, and 61.9%, respectively. The qualitative assessment of the model demonstrated its effectiveness under varying traffic scenarios and lighting conditions. Furthermore, it obtained an overall counting accuracy of 90%. These findings demonstrate the potential of the proposed framework for real-time traffic analysis and intelligent transportation systems.

@article{Shukla2026,
  title     = {HETRA: A Real-Time Traffic Detection and Directional Counting System Using YOLOv8 and DeepSORT},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {7},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Kuldeep Kumar Shukla and Ajai Kumar Singh},
  doi       = {10.14569/IJACSA.2026.0170725},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170725}
}