Automatic single-view character model reconstruction

Abstract

Figure 1: This paper presents a method for automatically constructing 3D meshes from a single piece of concept artwork. This figure shows how a typical 2D concept image (a) is separated from the background (b) and how a bent skeleton (c) is generated and used to create polygonal shells (d). The profile view (e) shows clearly how the shells create a 3D representation of the input image. The final model (f -h) is mapped with textures based on the initial concept artwork and contains bone and vertex weighting information appropriate for animation. Details such as the hair and ears are correctly transformed into 3D, and mechanical objects such as the gun and knife have sharp edges. AbstractIn this paper we present a new method for automatically constructing 3D meshes from a single input image. With the increasing content demands of modern digital entertainment and the expectation of involvement from users, automatic artist-free systems are an important step in allowing user generated content and rapid game prototyping. Our system proposes a novel heuristic for the creation of a 3D mesh from a single piece of non-occluding 2D concept art. By extracting a skeleton structure, approximating the 3D orientation and analysing line curvature properties, appropriate centrepoints can be found around which to create the cross-sectional slices used to build a final triangle mesh. Our results show that a single 2D input image can be used to generate a rigged 3D lowpolygon model suitable for use in realtime applications.

Gesture3D

“…As pointed out by Bessmeltsev et al [2015], modeling a posed character given a single contour sketch is inherently ambiguous, as the contours alone allow multiple perceptually plausible interpretations. To generate the user-desired models, existing methods either rely on incremental contour drawing order [Cherlin et al 2005], employ primitive-based annotation [Cherlin et al 2005;Gingold et al 2009;Shtof et al 2013], require users to provide a matching posed 3D skeleton , or enforce a range of strong simplifying assumptions [Cordier et al 2011;Buchanan et al 2013;Karpenko and Hughes 2006] -for instance, Cordier et al [2011] only handle mirror-symmetric poses, while other methods assume that the 3D generators of the drawn contours are planar and near-parallel to the view plane [Buchanan et al 2013;Karpenko and Hughes 2006;Bai et al 2016], an assumption that does not hold for most character poses . Our work addresses generic gesture drawings with no such limitations (Figure 1), solving an essentially inverse problem to that of : instead of computing a character geometry given a sketch and a 3D pose, we compute a character pose given a gesture drawing and a character model.…”

A Survey of Sketch Based Modeling Systems

“…The system then wraps the spine with the polygonal mesh and then uses constraint Delaunay triangulation of the polygon. Pruning of insignificant branches is performed using [Prasad 1997]. The algorithm progresses through several refinement steps to obtain a smooth and symmetric 3D surface.…”

“…Automatic single-view character model reconstruction: Buchanan et al [Buchanan et al 2013] have demonstrated a system which generates lowpolygon models directly from input character sketches without any interactivity by the artist. Their algorithm first computes a skeleton from the image and then draws cross section curves along the skeleton using a novel proposed heuristic.…”

“…This technique can be useful for rapidly generating sub parts of the human character, such as limbs and torso as these parts are not very complicated anatomically. Figure 5 shows an overview of the process in [Buchanan et al 2013]. [Buchanan et al 2013].…”

“…(Reproduced from [Buchanan et al 2013) 3.6. Geosemantic Snapping for Sketch-Based Modeling: In [Shtof et al, 2013] Shtof et al have introduced a snapping technique which automatically reshapes and snaps simple 3D geometric primitives to 2D sketch primitives, then improves the model globally by inferring geosemantic constraints that link the different parts.…”