Object localization on building plans using Machine Learning techniques
- Sponsored by: Unetiq GmbH
- DI Incubator: Startup Unetiq
- Project Leader: Dr. Ricardo Acevedo Cabra
- Scientific Lead: M.Sc Lucas Spreiter, Stefan Reuther
- Term: Winter semester 2019
Results of this project are explained in detail in the final documentation and presentation.
Unetiq is a Munich based start-up, that is developing the worlds first Artificial Intelligence architect. We use various technologies to automate architectural services, e.g. building equipment planning or building design. Our vision is to fully automate the architectural process and build the world’s first AI-designed building.
Currently, architects design building ground plans and hand these over to subcontractors, e.g. electrical engineers, to carry out specialized planning tasks. Ground plans are designed with Computer-Aided-Design (CAD) software. They are translated into a special file format that consists of basic geometric shapes such as splines, points and rectangles. To enable algorithms an understanding of the semantic structure of a given plan, contextual information about different rooms and objects is required, for example the position and size of a kitchen sink.
The goal of this project is to build an algorithm that enables the identification and localization of relevant objects in a given plan. This can be achieved using current state-of-the-art techniques such as Region Convolutional Neural Networks (R-CNN). We support students by providing expert knowledge in Deep Learning as well as cloud infrastructure by our service partners.