Abstract The road traffic accidents have remained a significant factor to death and economical losses across the world making it important to have the right and prompt accident detection systems. Manual monitoring with sensors or other traditional methods of monitoring tend to be limited regarding cost, scalability, and timeliness of response. This paper presents an artificial intelligence-based real-time road accident detection framework to solve these issues on the basis of video surveillance information. They use the deep convolutional neural networks (CNNs) with a single pipeline of preprocessing that includes image normalization, data augmentation and stratified dataset splitting. Curated collection of accident images is evaluated systematically through four known CNN architectures namely: VGG16, Inception V3, Mobile Net and DenseNet 121. Also, a new hybrid that integrates the power of VGG16 in the extraction of deep features and the efficiency of Mobile Net is presented. The experimental findings prove that the hybrid architecture performs better than single models with an accuracy of 95.60%, precision of 95.90%, a recall of 94.70%, and an F1-score of 95.30%. The results demonstrate the effectiveness, strength, and scalability of the suggested strategy, which makes it applicable to the smart transportation system and real-time applications in the context of traffic safety.
Wani et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: