SecondSkinby Chris De Paoli, Karan Singh



Computer Graphics and Computer-Aided Design


ACM Reference Format

De Paoli, C., Singh, K. 2015. SecondSkin: Sketch-Based Construction of Layered 3D Models.

ACM Trans. Graph. 34, 4, Article 126 (August 2015), 10 pages. DOI = 10.1145/2766948

Copyright Notice

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profi t or commercial advantage and that copies bear this notice and the full citation on the fi rst page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specifi c permission and/or a fee. Request permissions from

SIGGRAPH ?15 Technical Paper, August 09 ? 13, 2015, Los Angeles, CA.

Copyright 2015 ACM 978-1-4503-3331-3/15/08 ... $15.00.


SecondSkin: Sketch-based Construction of Layered 3D Models

Chris De Paoli Karan Singh

University of Toronto

Figure 1: 2D strokes sketched on and around 3D geometry form the input to SecondSkin (a). Layered structures are represented as solid models with volumes bounded by surface patches and curves (b). A majority (91%) of sketch strokes are perceived by viewers as one of four curve-types (c). We automatically classify these strokes based on the relationship between 2D strokes and underlying 3D geometry, producing 3D curves, surface patches, and volumes (d), resulting in layered 3D models suitable for conceptual design (e).


SecondSkin is a sketch-based modeling system focused on the creation of structures comprised of layered, shape interdependent 3D volumes. Our approach is built on three novel insights gleaned from an analysis of representative artist sketches. First, we observe that a closed loop of strokes typically define surface patches that bound volumes in conjunction with underlying surfaces. Second, a significant majority of these strokes map to a small set of curvetypes, that describe the 3D geometric relationship between the stroke and underlying layer geometry. Third, we find that a few simple geometric features allow us to consistently classify 2D strokes to our proposed set of 3D curve-types. Our algorithm thus processes strokes as they are drawn, identifies their curve-type, and interprets them as 3D curves on and around underlying 3D geometry, using other connected 3D curves for context. Curve loops are automatically surfaced and turned into volumes bound to the underlying layer, creating additional curves and surfaces as necessary. Stroke classification by 15 viewers on a suite of ground truth sketches validates our curve-types and classification algorithm. We evaluate SecondSkin via a compelling gallery of layered 3D models that would be tedious to produce using current sketch modelers.

CR Categories: I.3.5 [Computer Graphics]: Computational Geometry and Object Modeling?Geometric algorithms, and systems

Keywords: sketch-based modeling, layers, shells 1 Introduction

Current 3D conceptual design tools, regardless of being based on metaphors of sketching and sculpting or traditional CAD modeling, typically focus on the creation of the skin or visible surface of 3D objects. Physical objects, both organic and man-made, however, are often layered assemblies: comprising parts segmented by form, function or material, built over each other (Figure 1, 2 and 13).

While research in character skinning and animation has noted the importance of conceptual anatomic layers for a quarter century now [Chadwick et al. 1989], 3D conceptual design tools, to date, have largely ignored what lies beneath the skin. SecondSkin addresses this problem: the fluid sketch-based creation of layered 3D structures.

A defining aspect of layered modeling is the geometric dependence of layers on underlying layers. This is clearly evidenced in a multitude of books and tutorials on sketching and concept art [Davies and Scott 2012], where maquettes of underlying layers are used as a visual reference on and around which to draw subsequent layers (Figure 2). Prior work in sketch-based modeling [Kara and Shimada 2007; Nealen et al. 2007; Takayama et al. 2013], typically interprets sketched strokes as lying on template objects or the evolving 3D geometric skin of the object. In the context of layered modeling however, we expect that sketched strokes are largely drawn around underlying template objects, to build new layered structures.

While projecting a 2D stroke drawn from a given view on to 3D geometry is mathematically precise and straightforward, inferring a 3D curve from such a 2D stroke to lie around 3D geometry, is generally

ACM Transactions on Graphics, Vol. 34, No. 4, Article 126, Publication Date: August 2015

Figure 2: 2D concept art for armour variations drawn over a 3D character, c?Paul Richards. ill-defined. Despite this, design strokes representing layered geometry trigger 3D percepts that are consistently imagined by viewers (Figure 1(c)). Recently, the formulation of a number of perceived 3D relationships between connected 2D strokes have been exploited to lift design sketches off the page into 3D [Xu et al. 2014]. Complementary to these relationships and perhaps more important for layered modeling, are the perceived 3D relationships between a 2D stroke and the 3D geometry of the layer over which it is drawn. We discover through conversations with artists and analysis of layered sketches (Figures 1(b) and 10), that a majority of design strokes (286 of 313 strokes in Figure 1(c)),? 91%, are perceived in relation to the underlying layer geometry, as one of four 3D curve-types shown inset: shell contour, shell projection, tangent plane and normal plane. shel contour projection tangent normal

These curve types have individually seen use in sketching interfaces, for example, contours for model deformation [Zimmermann et al. 2007] or normal curves for 3D painting [Schmid et al. 2011]. Together however, they capture the bulk of design strokes drawn to depict layered structures. Strokes that do not belong to these categories, either do conform to them from a different viewpoint, or can be created from context, by anchoring them to strokes of the above curve-types interpreted in 3D. We further find that these 2D strokes can be robustly classified as one of the 3D curve-types, by simply observing the 2D spatial relationships of the curve relative to the underlying geometry.