Abstract:
This article presents a description of an online service for automated image-based evaluation of visual attractiveness of residential properties. The system allows to determine the room type in a photo, analyse and assess visual attractiveness both for specific rooms and for the whole real estate object based on training data.The work examines the issue of low digitalisation of the real estate sector. It also demonstrates the difficulties faced by estate agents, potential buyers and tenants on a daily basis, which confirms the necessity of developing such a system.The architecture of the developed system is described, including client-server interaction, neural network image processing pipeline, as well as mechanisms of integration with existing real estate platforms via REST API. Key functional and non-functional requirements to the system are reviewed to ensure high classification accuracy and fast image processing.The stages of informational interaction demonstrating the optimised business process with the application of the proposed solution are presented. A scenario of the service usage visualised with the help of a corresponding diagram is developed.
Keywords:automation, real estate market, neural network models, computer vision, evaluation of the appearance of premises.
Fulltext:
PDF file (432 kB)