Workflow:

In Virtual Rome 2.0 project, the input information is GIS based: the well-known DEM, raster and vector files, already described in the previous post, named "Virtual Rome 2.0 dataset". The final goal is to create realtime web applications and computer animation; actually in this project the focus is on the mentioned first goal. For reaching that, the input data has to be transformed, this is the landscape generation from GIS data. The other kind of input data are 3D models, such as trees, plants, constructions, buildings, that will be added to the 3D virtual environment in a further step of the process. Aswell as the procedural modelling, which briefly is the creation of 3D models and textures from a set of rules. In this project, the buildings are created in such way, and they are directly added into the realtime applications. For further information in procedural modelling, one can visit this link: http://www.cs.berkeley.edu/~sequin/PAPERS/Parish_Mueller_Cities.pdf 

Aim of this research:

We now focus on the landscape generation step, where the GIS data has to be converted into 3D textured terrain. At this point, the 3D models (trees, plants) are not taken into account, so we only need to generate the terrain and the soil textures. For this project, several softwares are intended to be tested, as said on the post "State of the Art": VNS, VUE, Terragen 2, and Blender. Currently, for other projects of the same aim of Virtual Rome 2.0, VNS is used for the landscape reconstruction of soil textures. It's proved to be a well adaptative software for GIS input data. For example, it allows to assign ecosystems directly to each color of the raster colormap, which improve noticeably the cost in time of the process. Also, WNS works with georefered data, which allows to keep the cartographic reference system taken from the GIS layers. Finally, other important feature for the project is that it exports the final image texture into several high resolution images, which are tiled and auto-georeferenced. However, the results of the landscape textures doesn't seem as realistic as other lanscape generators produce, like Vue Terragen 2. The graphics already seems a bit old fashioned, and 3DNature doesn't seem to release a new version of their software VNS3. Therefore, it's been considered to move to another software, or group of softwares/technology, searching for a new solution which guarantees better results with an affordable cost in both time and economy.