Figure 5
From: Scene Segmentation with Low-Dimensional Semantic Representations and Conditional Random Fields
![Figure 5](http://media.springernature.com/full/springer-static/image/art%3A10.1155%2F2010%2F196036/MediaObjects/13634_2010_Article_2698_Fig5_HTML.jpg)
Patch-level precision for various combinations of neighborhood sizes, global aggregate features, and spatial layout features. See the text for details.