Texture, highlights, and shading are some of many visual cues that allow humans to perceive material appearance in single pictures. Yet, recovering spatially-varying bi-directional reflectance distribution functions (SVBRDFs) from a image based on such has challenged researchers computer graphics for decades. We tackle lightweight capture by training deep neural network automatically extract make sense these cues. Once trained, our is capable per-pixel normal, diffuse albedo, specular albedo...

We describe a new vector-based primitive for creating smooth-shaded images, called the diffusion curve . A partitions space through which it is drawn, defining different colors on either side. These may vary smoothly along curve. In addition, sharpness of color transition from one side to other can be controlled. Given set curves, final image constructed by solving Poisson equation whose constraints are specified gradients across all curves. Like primitives, curves conveniently support...

For many computational photography applications, the lighting and materials in scene are critical pieces of information. We seek to obtain intrinsic images , which decompose a photo into product an illumination component that represents effects reflectance is color observed material. This under-constrained problem automatic methods challenged by complex natural images. describe new approach enables users guide optimization with simple indications such as regions constant or illumination....

True2Form is a sketch-based modeling system that reconstructs 3D curves from typical design sketches. Our approach to infer form 2D drawings novel mathematical framework of insights derived perception and literature. We note designers favor viewpoints maximally reveal shape information, strategically sketch descriptive convey intrinsic properties, such as curvature, symmetry, or parallelism. Studies indicate viewers apply these properties selectively envision globally consistent shape. mimic...

3D modeling remains a notoriously difficult task for novices despite significant research effort to provide intuitive and automated systems. We tackle this problem by combining the strengths of two popular domains: sketch-based procedural modeling. On one hand, exploits our ability draw but requires detailed, unambiguous drawings achieve complex models. other automates creation precise detailed geometry tedious definition parameterization Our system uses collection simple grammars, called...

Sketch-based modeling strives to bring the ease and immediacy of drawing 3D world. However, while drawings are easy for humans create, they very challenging computers interpret due their sparsity ambiguity. We propose a data-driven approach that tackles this challenge by learning reconstruct shapes from one or more drawings. At core our is deep convolutional neural network (CNN) predicts occupancy voxel grid line drawing. This CNN provides us with an initial reconstruction as soon user...

An intrinsic image is a decomposition of photo into an illumination layer and reflectance layer, which enables powerful editing such as the alteration object's material independently its illumination. However, decomposing single highly under-constrained existing methods require user assistance or handle only simple scenes. In this paper, we compute decompositions using several images same scene under different viewpoints lighting conditions. We use multi-view stereo to automatically...

Abstract Empowered by deep learning, recent methods for material capture can estimate a spatially‐varying reflectance from single photograph. Such lightweight is in stark contrast with the tens or hundreds of pictures required traditional optimization‐based approaches. However, image often simply not enough to observe rich appearance real‐world materials. We present deep‐learning method capable estimating variable number uncalibrated and unordered captured handheld camera flash. Thanks an...

This paper presents an interactive watercolor rendering technique that recreates the specific visual effects of lavis watercolor. Our method allows user to easily process images and 3d models is organized in two steps: abstraction step uniform color regions effect filters resulting abstracted image obtain watercolor-like images. In case environments we also propose methods produce temporally coherent animations keep a pigment repartition while avoiding shower door effect.

In this paper, we present a method for creating watercolor-like animation, starting from video as input. The involves two main steps: applying textures that simulate watercolor appearance; and simplified, abstracted version of the to which texturing operations are applied. Both these steps subject highly visible temporal artifacts, so primary technical contributions paper extensions previous methods abstraction provide coherence when applied sequences. To maintain textures, employ texture...

Vector drawing is a popular representation in graphic design because of the precision, compactness and editability offered by parametric curves. However, prior work on line vectorization focused solely faithfully capturing input bitmaps, largely overlooked problem producing compact editable curve network. As result, existing algorithms tend to produce overly-complex drawings composed many short curves control points, especially presence thick or sketchy lines that yield spurious at...

We facilitate the creation of 3D-looking shaded production drawings from concept sketches. The key to our approach is a class commonly used construction curves known as cross-sections , that function an aid both sketch and viewer understanding depicted 3D shape. In particular, intersections these curves, or cross-hairs convey valuable information, viewers compose into mental model overall sketch. use artist-drawn automatically infer normals across sketch, enabling 3D-like rendering....

We introduce a method to compute intrinsic images for multiview set of outdoor photos with cast shadows, taken under the same lighting. use an automatic 3D reconstruction from these and sun direction as input decompose each image into reflectance shading layers, despite inaccuracies missing data model. Our approach is based on two key ideas. First, we progressively improve accuracy parameters our formation model by performing iterative estimation combining lighting simulation 2D optimization...

Product designers extensively use sketches to create and communicate 3D shapes thus form an ideal audience for sketch-based modeling, non-photorealistic rendering sketch filtering. However, sketching requires significant expertise time, making design a scarce resource the research community. We introduce OpenSketch , dataset of product aimed at offering rich source information variety computer-aided tasks. contains more than 400 representing 12 man-made objects drawn by 7 15 varying...

Abstract Creating a virtual city is demanded for computer games, movies, and urban planning, but it takes lot of time to create numerous 3D building models. Procedural modeling has become popular in recent years overcome this issue, creating grammar get desired output difficult consuming even expert users. In paper, we present an interactive tool that allows users automatically generate such from single image building. The user selects photograph highlights the silhouette target as input our...

CAD modeling, despite being the industry-standard, remains restricted to usage by skilled practitioners due two key barriers. First, user must be able mentally parse a final shape into valid sequence of supported commands; and second, sufficiently conversant with software packages execute corresponding commands. As step towards addressing both these challenges, we present Free2CAD wherein can simply sketch our system parses input strokes commands expressed in simplified language. When...

Visual content is often better communicated by simplified or exaggerated images than the "real world like" images. In this paper, we offer a tool for creating such enhanced representations of photographs in way consistent with original image content. To do so, develop method to identify relevant structures and their importance. Our approach (a) uses edges as basic structural unit image, (b) proposes tools manipulate structure flexible way, (c) employs gradient domain processing techniques...

In this paper, we present a method for creating watercolor-like animation, starting from video as input. The involves two main steps: applying textures that simulate watercolor appearance; and simplified, abstracted version of the to which texturing operations are applied. Both these steps subject highly visible temporal artifacts, so primary technical contributions paper extensions previous methods abstraction provide coherence when applied sequences. To maintain textures, employ texture...

Intrinsic images aim at separating an image into its reflectance and illumination components to facilitate further analysis or manipulation. This separation is severely ill posed the most successful methods rely on user indications precise geometry resolve ambiguities inherent this problem. In paper, we propose a method estimate intrinsic from multiple views of outdoor scene without need for with few manual steps calibrate input. We use multiview stereo automatically reconstruct 3D point...

We present an interactive drawing tool that provides automated guidance over model photographs to help people practice traditional drawing-by-observation techniques. The literature describes a number of techniques %support this task and gain consciousness the shapes in scene their relationships. compile these derive set construction lines we automatically extract from photograph. then display guide its manual reproduction by user on canvas. Finally, use shape-matching register user's sketch...

We present a thorough study to evaluate different light field editing interfaces, tools and workflows from user perspective. This is of special relevance given the multidimensional nature fields, which may make common image tasks become complex in space. additionally investigate potential benefits using depth information when editing, limitations imposed by imperfect reconstruction current techniques. perform two experiments, collecting both objective subjective data varied number increasing...

Designers frequently draw curvature lines to convey bending of smooth surfaces in concept sketches. We present a method extrapolate rough sketch, recovering the intended 3D field and surface normal at each pixel sketch. This information allows enrich sketch with 3D-looking shading texturing. first introduce regularized that model designers over curved surfaces, encompassing their extension as geodesics flat or umbilical regions. build on this define orthogonal cross assigns two point...

We present CASSIE, a conceptual modeling system in VR that leverages freehand mid-air sketching, and novel 3D optimization framework to create connected curve network armatures, predictively surfaced using patches with C0 continuity. Our provides judicious balance of interactivity automation, providing homogeneous drawing interface for mix curves, networks, surface patches. encourages aids users consistent networks easing the transition from ideation concept modeling. A comprehensive user...

Abstract Recovering spatially‐varying materials from a single photograph of surface is inherently ill‐posed, making the direct application gradient descent on reflectance parameters prone to poor minima. Recent methods leverage deep learning either by directly regressing using feed‐forward neural networks or latent space SVBRDFs encoder‐decoder generative adversarial followed gradient‐based optimization in space. The former fast but does not account for likelihood prediction, i.e ., how well...

For many computational photography applications, the lighting and materials in scene are critical pieces of information. We seek to obtain intrinsic images, which decompose a photo into product an illumination component that represents effects reflectance is color observed material. This under-constrained problem automatic methods challenged by complex natural images. describe new approach enables users guide optimization with simple indications such as regions constant or illumination....