Efficient Road Traffic Video Congestion Classification Based on the Multi-Head Self-Attention Vision Transformer Model

Due to rapid population growth, traffic congestion has become one of the major issues in urban areas. The utilization of technology may help to address this issue. This paper proposes a new Multi-head Self-attention Vision Transformer (MSViT) based macroscopic approach, for road traffic congestion classification. To evaluate this approach, we use the UCSD (University of California San Diego) dataset that includes different weather conditions (clear, overcast and rainy) and different traffic scenarios (light, medium and heavy). The classification accuracy reached a high level of 99.76% with this dataset and 99.37% when night-mode frames are added to it. The proposed MSViT based method outperforms the state-of-the-art macroscopic and microscopic methods that have been evaluated using the same UCSD dataset, which makes it an efficient solution for traffic congestion prediction.