Streetlight inventory of a city aid in evaluating the road safety and will contribute in the evaluations of various other metrics associated with the urban planning. The traditional approaches to create the inventory data is time-consuming and resource intensive. To address this gap, a scalable approach to create streetlight inventory using google street view data and deep learning-based object detection is proposed in this study. Street view data at various inclinations including 0°, 60°, 120°, 180, 240, and 300°, ensuring full 360° coverage at a location is collected. Augmentations are further applied for these datasets to make the dataset robust. Later, the obtained imagery is used for manual annotation with the help of LabelMe, by drawing bounding box coordinates aligning to the YOLOv5 annotation scheme. Subsequently, the annotated data is used for training and the developed models is used to evaluate the presence of street lights within each of the accessed imagery using deep learning model built on the framework of YOLOv5. The proposed approach is distinct from the scholarly available in terms of considering the 360 degree data a location. The entire framework is demonstrated considering Nellore City, Andhra Pradesh, India as a study region. A total of 4,418 panoramas have been fetched and the proposed farmework has been applied to detect the street lights. Findings suggested that the proposed approach with an average precision at IoU threshold 0.5 (AP@0.5) of 0.804, an F1-score of 0.81, recall of 0.774, and the shortest inference time of 1.8 ms per image, demonstrates its suitability for large-scale implementation. The proposed algorithm can be extended to any of the locations for creating a repository of street lights within a city.