Single-View 3D Reconstruction via Differentiable Rendering and Inverse Procedural Modelingстатья
Информация о цитировании статьи получена из
Scopus
Статья опубликована в журнале из списка Web of Science и/или Scopus
Дата последнего поиска статьи во внешних источниках: 10 апреля 2024 г.
Аннотация:Three-dimensional models, reconstructed from real-life objects, are extensively used invirtual and mixed reality technologies. In this paper we propose an approach to 3D model reconstruc-tion via inverse procedural modeling and describe two variants of this approach. The first option isto fit a set of input parameters using a genetic algorithm. The second option allows us to significantlyimprove precision by using gradients within the memetic algorithm, differentiable rendering, anddifferentiable procedural generators. We demonstrate the results of our work on different models,including trees, which are complex objects that most existing methods cannot reconstruct. In ourwork, we see two main contributions. First, we propose a method to join differentiable renderingand inverse procedural modeling. This gives us the ability to reconstruct 3D models more accuratelythan existing approaches when few input images are available, even for a single image. Second, wecombine both differentiable and non-differentiable procedural generators into a single frameworkthat allows us to apply inverse procedural modeling to fairly complex generators. We show that bothvariants of our approach can be useful: the differentiable one is more precise but puts limitations onthe procedural generator, while the one based on genetic algorithms can be used with any existinggenerator. The proposed approach uses information about the symmetry and structure of the objectto achieve high-quality reconstruction from a single image