Automatic semantic analysis of 3D content in digital repositories

The increasing popularity of 3D technologies is having an impact on the amount of content that is being produced by users of these technologies. Witnessing the explosion of content such as images, music and videos available on the web, it is not difficult to predict that 3D will be the next type of content to undergo this effect. The research community has been taking action to ensure 3D content can be stored and managed in databases or repositories in order to be accessible to a wide variety of users.

Nevertheless, searching for 3D content in these repositories is not an easy task. The main problem is that although a digital 3D representation of a physical object is a more accurate representation, the way that the information is stored means that automated solutions for understanding what the content represents is an unsolved challenged.

To address this problem, the research community has created ways to tag or 'attach' additional information to the 3D content, as is done with 2D images, to support the computer's understanding of what the 3D content represents. However, this process is currently slow as it relies on mostly manual or semi-automatic techniques. This project will take these basic techniques forward by researching state of the art mechanisms to automate the enrichment of 3D content.

Project timeframes

The project ran from August 2014 to May 2016.

Project aims

This project aims to demonstrate the use of 3D technologies for documenting and analysing shape in the cultural heritage domain. This will be done by focusing on Cultural Heritage artefacts, in particular Regency architectural ornamental artefacts, to understand how the shape of an artefact might provide us with information about it (e.g. its origin, artistic style, production methods).

It is currently very challenging to infer this high level information automatically. The project will thus combine expertise in shape analysis, the semantic web and Cultural Heritage in order to develop innovative techniques to automatically understand what the 3D content might represent. This process is referred to as "automatic semantic enrichment" and will allow the 3D content to be linked to a vast amount of information and knowledge which will facilitate making connections with other pieces of information.

Project findings and impact

As a result of this project, we expect searching for the most relevant item of 3D content amongst the petabytes of information stored in the database will be considerably improved. In turn, this will improve the availability and use of 3D content for different purposes. For instance, the project will demonstrate how the research can support the restoration of historical buildings.

Research team

Dr K Rodriguez Echavarria

Dr Ran Song

Dean Few

Output

Karina Rodriguez Echavarria and Ran Song (2015). Studying Shape Semantics of an Architectural Moulding Collection: Classifying Style Based on Shape Analysis Methods. In Proceedings of the 2nd International Congress on Digital Heritage 2015. Granada, Spain.
Best paper award in the Analysis and Interpretation track

Partners