The advancement in the field of Urban Building Energy Modelling (UBEM) is assisting urban planners and managers to design and operate cities to meet environmental UBEM depends upon the quality and level of details (LoD) of the inputs to the model. The inadequacy and quality of relevant input data pose challenges. This paper analyses the usefulness of different methodologies for developing a 3D building stock model of Ahmedabad, India, recognizing data gaps and heterogenous development of the city over time. It evaluates the potentials, limitations, and challenges of remote sensing techniques namely (a) Satellite imagery (b) LiDAR and (c) Photogrammetry for this application. Further, the details and benefits of data capturing through UAV assisted Photogrammetry technique for the development of the 3D city model are discussed. The research develops potential techniques for feature detection and model reconstruction using Computer vision on the Photogrammetry reality mesh. Preliminary results indicate that the use of supervised learning for Image based segmentation on the reality mesh detects building footprints with higher accuracy as compared to geometry-based segmentation of the point cloud. This methodology has the potential to detect complex building features and remove redundant objects to develop the semantic model at different LoDs for urban simulations. The framework deployed and demonstrated for the part of Ahmedabad has a potential for scaling up to other parts of the city and other Indian cities having similar urban morphology and no previous data for developing a UBEM.