The Polymathic Machine project (French FUI 2011-2014) aims to develop a platform for storing, sharing, indexing and searching (including by content) very large collections of so-called graphical assets (e.g., images, textures, movements). The project gathers game companies (Bulkypix and Kiloton), Motion Capture companies Mocaplab, and the Cnam R&D institute4 in charge of designing the platform architecture and testing the proposed solution. In brief, the motivation of the envisioned system is to share the assets produced by game companies in a common market- place, in order to leverage design efforts and make their results reusable by other graphic artists and developers. In this vision, a company C aiming at creating a new game should be able to search the marketplace for assets representing objects (some furniture from the Roman Empire), characters (the Red Knight with his splendid armor), scenes (a typical landscape from Orion III), or even low-level graphical components such as textures, motion models, etc. Assets of interest can then be bought and incorporated in the new game by C, thereby minimizing its development efforts. In addition to game and motion capture companies, our solution provides an open market- place where creatives can produce and market their assets, much as “App Stores” have opened the marketplace to independent software developers.
The platform’s goals give rise to several advanced requirements regarding the management of assets. Some are highly specific to the nature of the data at hand. For instance, content-based search (e.g., find some furniture based on its shape and color) requires the production (and indexing) of descriptors for multimedia contents supporting similarity search. This covers 2D descriptors for images, 3D descriptors for objects and scenes, and even motion descriptors. Other requirements pertain to the general issue of managing a shared repository of complex documents, fed by individual contributors working from their individual and familiar desktop environments. In the envisioned scenario, assets are created and updated locally with some task-specific desktop software (e.g., Photoshop), and must then be transferred to the marketplace’s repository. Such a transfer involves copyright tagging, format conversion, descriptor extraction, and generally requires a tight connection between each user’s desktop and the shared repository. A major goal of the project was to make this connection as easy as possible in order to avoid the burden of a complex procedure that periodically collects, transforms and transfers recently created or updated assets. Graphic artists should be able to focus on the asset creation process in a familiar environment, with limited additional constraints, and the connection with the shared marketplace should be handled as transparently as possible.
Links: Cnam’s demonstration
[tplist author=44 order=”year DESC” tag=32]