Estimation of building parameters from street view images

Buildings are important parts of urban area. Their parameters are useful for urban planning, house reconstruction, and navigation. Apart from the aerial images that acquire the information from above, street view images have been identified as an emerging data source to estimate building parameters. A proof-of-concept research is conducted to estimate firstly the building types from street view images. In this work, we proposed an end-to-end approach for building type classification. The street view images are downloaded from Google Street View based on OSM road networks. An automatic labelling process was proposed to prepare a training dataset. Using the state-of-art Convolutional Neural Networks. Building types of each building instances on the OSM can be classified based on the interpretation of street view image.