Wednesday 3 March 2010

Problems & Solutions


Communication:

Communication was one of the first problems we had as a group, however we quickly remedied this problem by exchanging mobile phone numbers and e-mails. W also decided to use online messaging applications such as MSN & Skype.

Texturing:

We have had some issues regarding textures when attempting to import a scene from another computer. This seems to be a rather annoying and frequent problem. One way we have solved this is to manually re-assign the textures, but this takes up valuable time. The other day Dan showed me and Tom a way to set the textures by setting the project scene and its textures under source images rather than documents. I have to say that I do not yet fully understand how the texture system works. I also understand that other students have been experiencing similar problems.

Rendering:

The other problem is that some of our scenes are far too complex for our laptops to render in terms of polygons and information. Sometimes it is due to a lack of memory or RAM. However we have solved this problem by dismissing the fly-through approach and exploring the environment by using numerous shots. Thus we do not have to render all the components within a single shot. We have also taken this a stage further by making sure that any environment components not visible within the camera frame are deleted to cut down render time and lag/ delay issues. There is no point in including elements which the viewer cannot see. This is precisely what they do in film, however from the experience on this project, I have realised that using this technique is even more important when it comes to CG 3D environments.
We have discovered that we cannot use the render farm effectively as Qube is only compatible with mental ray and only works with .tiff format textures. We have used many .jpg when texturing, and it would be a painstakingly slow process of re-formating hundreds of textures. There were literally thousands of errors that we would have to had to address. It has also come to my attention that during the last weeks Qube buckled under the pressure of too many projects being uploaded to be rendered which exceeded the memory limit. Numerous projects which were 4 Gb and over crashed the render farm, thus this is another reason why it was a good decision to render locally on our own laptops. Due to these problems we as a group made a conscious decision to render out each scene using our own laptop computers.
We have also decided not to render our scenes with an occlusion layer as we discovered that rendering this layer almost doubled render time. The result did not look too different from the
renders without an occlusion layer. The only way we could include an occlusion layer was by either rendering different layers separately of the same scene on several laptops simultaneously or the the render farm Qube was an option however I have already explained the issue with this method. So we decided to render without an occlusion render in order to cut down on render time.

No comments:

Post a Comment