Vehicle detection and surveillance of traffic are important areas of city planning, especially in highly populous countries with heavy road traffic like India. This study focused on training YOLO11 and YOLO12 to enhance the vehicle detection for the Indian traffic scenario and incorporating SAHI inference technique for improved detection of small-sized vehicles. Since Indian traffic is composed of diverse vehicles types like two-wheelers, auto-rickshaws, trucks, buses, tempos and cars, this research aims to achieve high accuracy and efficiency in vehicle detection. Performance of both the models was compared based on metrics like precision, recall, f1-score and speeds of inference, preprocessing and postprocessing. Experiments demonstrated how SAHI inference enhanced the small-size vehicle detection accuracy. Comparing both models showcased the performance trade-offs where YOLO12 performed better in terms of accuracy, whereas YOLO11 came out to be a more efficient model. While YOLO11 recorded minor misclassifications, YOLO12 showed no such hallucinations. These findings demonstrated the robustness of YOLO models when applied for complex challenges like Indian Traffic Vehicle detection.