In the developing countries like India, traffic maintenance on roads is some what crucial task. Vehicle detection plays an important role and is the basic step for automatic monitoring of traffic. Vehicle detection is the procedure that recognizes vehicles from image or video acquired by cameras. It has an important role in a variety of applications, including traffic monitoring, surveillance, self-driving cars, and intelligent transportation systems. The emergence of deep learning (DL) models with convolution neural networks (CNN) have shown a path to the vehi- cle detection problem. Several DL models with CNNs have been presented in the literature to tackle the problem of vehicle detection. All these models were not up to the mark to produce results especially for Indian scenario due to hectic traf- fic conditions. To address this, a new vehicle detection method using enhanced neural architecture search for Indian scenario is presented in this paper. The experimental findings show that, on the FGVD and IRUVD datasets, the proposed model outperforms YOLO NAS by 10.94% and 1.29%, respectively, with precision rates of 84.51% and 96.61%.