SMU Guildhall Thesis
The scene was constructed for my thesis while at the SMU Guildhall.
The thesis test aims to categorize the importance of different texture maps loaded onto a material. The scene showcases a variety of surface types to assist with the test.
The thesis test categorizes the importance of each texture map for every asset. The assets get assigned a hierarchy of what textures best convey the assets surface properties. Based on this hierarchy, each asset get grouped together on a single UV texture sheet. By prioritizing the grouping of the assets UVs, more visual control can be had during optimization.
For example, a backpack may receive a diffuse, specualr and normal map, each at the size of 512×512. However, most of the visual information is happening within the normal and specular maps. The backpack’s diffuse consists of very large color groups, that texture could take a huge reduction in size and still convey the needed surface information when using the regular specualr and normal maps. Once the scene was grouped together using the thesis test’s hierarchy, a sizable amount of space was saved as visual standards stayed nearly the same during optimization.
My overall goal with the theses was to find more ways to save on texture size stored in memory. During my research i came across Substance Designer and I began exploring its tools. I ended up reconstructed the thesis scene using SD materials while taking advantage of its node based texturing system and the expose function to maintain more control when back in UDK. The combination of the asset grouping and Substance designer, allowed for a substantial reduction in memory size while allowing for, in-game optimization and visual control.
Below are a few images form the second half of my thesis test. I displayed the textures on cubes as-well as on the assets for the hierarchy of texture maps. I did this to ensure that testers did not rely on a asset’s mesh to help identify its surface properties.