Банс Самюел Джордж -
Architectural Building Style Detection and Real Estate Classification Using a Tensorflow Framework
Системы больших данных
A pipeline is presented demonstrating how businesses can extract meaningful data from images to improve pricing models and enrich existing data. Using the example of building image data, images are collected and then processed at scale using a Faster-RCNN algorithm to infer architectural styles. This data is then mapped to produce an architectural mapping of London, as well as for tagging individual buildings within a real estate database with their architectural class. This paper demonstrates how this additional information could be utilised by a real estate firm, to provide greater controls over feature filtering, and thus present an enhanced user experience for individuals looking to sort properties based on styles and features that would traditionally be embedded in the image data.