Very large polygonal models are frequently used in scientific computing, mechanical engineering, or virtual medicine. An effective technique to handle these large models is occlusion culling. Like most large model techniques, occlusion culling trades overhead costs with the rendering costs of the possibly occluded geometry. In this paper, we present occlusion culling techniques for OpenSG. These techniques utilize OpenGL or OpenGL extensions to determine occlusion information. Techniques utilizing image space and temporal coherence are used to optimize occlusion query efficiency. Only the standard OpenSG data structures are used and no precomputation is necessary for the presented techniques.