Distributed, workflow-driven rendering of 3D object scenes on a big data processing platform

This paper aims to explore and experiment the use of a general purpose, distributed processing solution for the rendering of 3D object scenes. Although originally designed for the processing of Earth Observation data, the platform was built with flexibility and extensibility in mind. At its core, it is based on the principle of task modularity, aiming to both simplify the specification of processing algorithms and decrease their overall execution time. To this end, it relies on the dynamic generation and distribution of sub-tasks within a distributed processing environment. Throughout the paper, we will be demonstrating its potential for rendering both still frames and videos of complex object scenes, by setting up, in effect, an ad-hoc rendering farm. This will help to highlight both the platform's utility in exploiting inherent task and data parallelism in order to improve execution time and its great flexibility and adaptability to different usage scenarios, and to a variety of processing operations.