A 3D modeling methodology based on a concavity-aware geometric test to create 3D textured coarse models from concept art and orthographic projections. (November 2018)

• Record Type: Journal Article
• Authors: Silva Junior, Sergio N.; Chamone, Felipe C.; Ferreira, Renato C.; Nascimento, Erickson R.
• Is Part Of: Computers & graphics. Volume 76(2018)
• Page Start: 73
• Page End: 83
• Publication Date: 2018-11
• Subjects: 3D Modeling -- Voxel carving -- Texture generation; Computer graphics -- Periodicals; 006.6
• DOI: 10.1016/j.cag.2018.09.002
• Languages: English
• ISSNs: 0097-8493

Abstract

Highlights: A methodology for automatic 3D modeling from concept art is proposed. In addition, it provides a diffuse texture map based on input pieces of art colors. The algorithm keeps the 3D artist workflow by using three viewpoints of the object. Performing a geometric consistency overcomes the base algorithm drawbacks. Geometric consistency shows better results in sketches than Photo-consistency. Graphical abstract: Abstract: Creating textured 3D meshes of objects for real-time applications can be a laborious, slow and expensive task, demanding specific, highly specialized human resources such as 2D and 3D artists. In this paper, we present a fully automatic 3D modeling methodology based on silhouette carving, capable of creating textured 3D meshes from three pieces of concept art. Our method takes a set of the target object concept art in different views as input and generates a coarse 3D mesh alongside a diffuse color map for texturing the model. The coarse mesh is intended to replace the initial primitive mesh used on the modeling technique known as Box Modeling to accelerate the whole 3D model production. Although, in current 3D model production pipeline there are some more sophisticated methods based on sculpting and retopology, Box Modeling is still a heavily adopted technique used for man-made objects that do not require organic modeling. Our experiments show that it speeds up the 3D content production time up to 40% by providing the coarse model automatically. Also, our method preserve the artist's trace and can create more accurate meshes compared to a similar approach, photoconsistency-based, and learning-based methods.