3DScope: Detecting 3D Model Clones
Algorithm for extracting 3D models, indexing them using a value-insensitive normalization algorithm, and comparing the model indexes to detect cloned 3D models.
The unauthorized copying of 3D models robs designers of property, reduces innovation going into the creation of 3D models, creates liability for entities unknowingly using copied models, and indicates a lapse in IP security. With the vast amount of 3D assets in use, it is infeasible for humans to identify when their assets have been misappropriated by others. As use of 3D models for designs, gaming, metaverse, VR, animation and additive manufacturing applications increases, the need to identify stolen model assets becomes greater. This is a need for individual designers, corporations who utilize those assets, and software companies on the cutting edge of the industry. Current 3D model retrieval and identification approaches can be fooled by file recompression, asset compression and stretching, along with other file modifications.
OSU researchers have developed a method and software, 3DSCOPE, which detects 3D model clones and copies. This method is a value insensitive approach which avoids the pitfalls of comparing the hash of 3D files and other methods easily tricked by reformatting, model resizing, and file modifications. This approach is able to detect model copies even when raw data is altered, metadata is scrubbed, assets are rescaled, keyframes are dropped, and when other noise or reformatting changes asset values. By avoiding the pitfalls of hashing, and local extrema methods, 3DSCOPE detects clones in a precise, computationally efficient, and scalable manner.