We describe a novel method for searching 3D model collections using free-form sketches within a virtual environment as queries. As opposed to traditional sketch retrieval, our queries are drawn directly onto an example model. Using immersive virtual reality the user can express their query through a sketch that demonstrates the desired structure, color and texture. Unlike previous sketch-based retrieval methods, users remain immersed within the environment without relying on textual queries or 2D projections which can disconnect the user from the environment. We perform a test using queries over several descriptors, evaluating the precision in order to select the most accurate one. We show how a convolutional neural network (CNN) can create multi-view representations of colored 3D sketches. Using such a descriptor representation, our system is able to rapidly retrieve models and in this way, we provide the user with an interactive method of navigating large object datasets. Through a user study we demonstrate that by using our VR 3D model retrieval system, users can perform search more quickly and intuitively than with a naive linear browsing method. Using our system users can rapidly populate a virtual environment with specific models from a very large database, and thus the technique has the potential to be broadly applicable in immersive editing systems.
An early step in bottom-up diagram recognition systems is grouping ink strokes into shapes. This paper gives an overview of the key literature on automatic grouping techniques in sketch recognition. In addition, we identify the major challenges in grouping ink into identifiable shapes, discuss the common solutions to these challenges based on current research, and highlight areas for future work.
We present a novel context-based sketch classification framework using relations extracted from scene images. Most of existing methods perform sketch classification by considering individually sketched objects and often fail to identify their correct categories, due to the highly abstract nature of sketches. For a sketched scene containing multiple objects, we propose to classify a sketched object by considering its surrounding context in the scene, which provides vital cues for alleviating its recognition ambiguity. We learn such context knowledge from a database of scene images by summarizing the inter-object relations therein, such as co-occurrence, relative positions and sizes. We show that the context information can be used for both incremental sketch classification and sketch co-classification. Our method outperforms a state-of-the-art single-object classification method, evaluated on a new dataset of sketched scenes.
Complex vector drawings serve as convenient and expressive visual representations, but they remain difficult to edit or manipulate. For clean-line vector drawings of smooth organic shapes, we describe a method to automatically extract a layered structure for the drawn object from the current or nearby viewpoints. The layers correspond to salient regions of the drawing, which are often naturally associated to 'parts' of the underlying shape. We present a method that automatically extracts salient structure, organized as parts with relative depth orderings, from clean-line vector drawings of smooth organic shapes. Our method handles drawings that contain complex internal contours with T-junctions indicative of occlusions, as well as internal curves that may either be expressive strokes or substructures. To extract the structure, we introduce a new part-aware metric for complex 2D drawings, the radial variation metric, which is used to identify salient sub-parts. These sub-parts are then considered in a priority-ordered fashion, which enables us to identify and recursively process new shape parts while keeping track of their relative depth ordering. The output is represented in terms of scalable vector graphics layers, thereby enabling meaningful editing and manipulation. We evaluate the method on multiple input drawings and show that the structure we compute is convenient for subsequent posing and animation from nearby viewpoints.
We present a new approach to reconstruction of high-relief surface models from hand-made drawings. Our method is tailored to an interactive modeling scenario where the input drawing can be separated into a set of semantically meaningful parts of which relative depth order is known beforehand. For this kind of input, our technique allows inflating individual components to have a semi-elliptical profile, positioning them to satisfy prescribed depth order, and providing their seamless interconnection. Compared to previous methods, our approach is the first that formulates this reconstruction process as a single non-linear optimization problem. Because its direct optimization is computationally challenging, we propose an approximate solution which delivers comparable results orders of magnitude faster enabling an interactive user workflow. We evaluate our approach on various hand-made drawings and demonstrate that it provides state-of-the-art quality in comparison with previous methods which require comparable user intervention.
We present a framework for automatically creating a type of artwork in which 2D and 3D contents are mixed within the same composition. These artworks create plausible effects for the viewers by showing a different relationship between 2D and 3D at each viewing angle. As the viewing angle is changed, we can clearly see 3D elements emerging from the scene. When creating such artwork, we face several challenges. The main challenge is to ensure the continuity between the 2D and the 3D parts in terms of geometry and colors. We provide a 3D synthetic environment in which the user selects the region of interest (ROI) from a given scene object to be shown in 3D. Then we create a flat rendering grid that matches the topology of the ROI and attach the ROI to the rendering grid. Next we create textures for the flat part and the ROI. To enhance the continuity between the 2D and the 3D parts of the object, we include bas-relief profiles around the ROI. Our framework can be used as a tool in order to assist artists in designing such sculpture paintings. Furthermore, it can be applied by amateur users to create decorative objects for exhibitions, souvenirs, and homes.
We present an implicit approach for constructing smooth 3-D inscribed volumes intended for modeling porous structures, such as volcanic rocks, foam, radiolarians, and Swiss cheese. Positive inscribed volumes can model natural pebbles, and negative volumes can model porous structures. We introduce two techniques for blending and creating interconnections between these inscribed volumes to adapt our approach to both regular and irregular. We begin with a set of convex polytopes such as 3-D Voronoi diagram cells and compute inscribed volumes bounded by the cells. The cells can be irregular in shape, scale, and topology, and this irregularity transfers to the inscribed volumes, producing natural-looking spongy structures. Describing the inscribed volumes with implicit functions gives us the freedom to exploit volumetric surface combinations and deformation operations effortlessly.
The human figure is important in art. I discuss examples of the abstract depiction of the human figure and the challenge faced in algorithmically mimicking what human artists can achieve. The challenge lies in the human brain having enormous knowledge about the world and an ability to make fine distinctions about other humans from posture, clothing and expression. This allows a human to make assumptions about human figures from a tiny amount of data, and allows a human artist to take advantage of this when creating art. We look at examples from impressionist painting, cross-stitch, knitting, pixelated renderings in early video games, and the stylisation used by the artists of children's books.
We propose a framework for expressive non-photorealistic rendering of 3D computer graphics: MNPR. Our work focuses on enabling stylization pipelines with a wide range of control, thereby covering the interaction spectrum with real-time feedback. In addition, we introduce control semantics that allow cross-stylistic art-direction, which is demonstrated through our implemented watercolor, oil and charcoal stylizations. Our generalized control semantics and their style-specific mappings are designed to be extrapolated to other styles, by adhering to the same control scheme. We then share our implementation details by breaking down our framework and elaborating on its inner workings. Finally, we evaluate the usefulness of each level of control through a user study involving 20 experienced artists and engineers in the industry, who have collectively spent over 245 hours using our system. MNPR is implemented in Autodesk® Maya® and open-sourced through this publication, to facilitate adoption by artists and further development by the expressive research and development community.
One of the qualities sought in expressive rendering is the 2D impression of the resulting style, called flatness. In the context of 3D scenes, screen-space stylization techniques are good candidates for flatness as they operate in the 2D image plane, after the scene has been rendered into G-buffers. Various stylization filters can be applied in screen-space while making use of the geometrical information contained in G-buffers to ensure motion coherence. However, this means that filtering can only be done inside the rasterized surface of the object. This can be detrimental to some styles that require irregular silhouettes to be convincing. In this paper, we describe a post-processing pipeline that allows stylization filters to extend outside the rasterized footprint of the object by locally "inflating" the data contained in G-buffers. This pipeline is fully implemented on the GPU and can be evaluated at interactive rates. We show how common image filtering techniques, when integrated in our pipeline and in combination with G-buffer data, can be used to reproduce a wide range of "digitally-painted" appearances, such as directed brush strokes with irregular silhouettes, while keeping enough motion coherence.
We present the first empirical study on using color manipulation and stylization to make surgery images more palatable. While aversion to such images is natural, it limits many people's ability to satisfy their curiosity, educate themselves, and make informed decisions. We selected a diverse set of image processing techniques, and tested them both on surgeons and lay people. While many artistic methods were found unusable by surgeons, edge-preserving image smoothing gave good results both in terms of preserving information (as judged by surgeons) and reducing repulsiveness (as judged by lay people). Color manipulation turned out to be not as effective.
We introduce a novel approach that uses a generative adversarial network (GAN) to synthesize realistic oil painting brush strokes, where the network is trained with data generated by a high-fidelity simulator. Among approaches to digitally synthesizing natural media painting strokes, physically based simulation produces by far the most realistic visual results and allows the most intuitive control of stroke variations. However, accurate physics simulations are known to be computationally expensive and often cannot meet the performance requirements of painting applications.
In our work, we propose to replace the expensive fluid simulation with a neural network. The network takes the existing canvas and a new stroke trajectory as input and produces the height and color of the new stroke as output. We train the network with a dataset generated with a high quality offline simulator. The network is able to produce visual quality comparable to the offline simulator with better performance than the existing real-time oil painting simulator. Finally, we implement a real-time painting system using the trained network.
Digital media allows artists to create a wealth of visually-interesting effects that are impossible in traditional media. This includes temporal effects, such as cinemagraph animations, and expressive fluid effects. Yet these flexible and novel media often require highly technical expertise, which is outside a traditional artist's skill with paintbrush or pen. Fluid Brush acts a form of novel, digital media, which retains the brush-based interactions of traditional media, while expressing the movement of turbulent and laminar flow. As a digital media controlled through a non-technical interface, Fluid Brush allows for a novel form of painting that makes fluid effects accessible to novice users and traditional artists. To provide an informal demonstration of the medium's effects, applications, and accessibility, we asked designers, traditional artists, and digital artists to experiment with Fluid Brush. They produced a variety of works reflective of their artistic interests and backgrounds.
We present a computational framework for creating swept volume light painting and kinetic photography. Unlike conventional light painting technique using hand-held point light source or LED arrays, we move a flat-panel display with robot in a curved path. The display shows real-time rendered contours of a 3D object being sliced by the display plane along the path. All light contours are captured in a long exposure and constitute the virtual 3D object augmented in the real space. To ensure geometric accuracy, we use hand-eye calibration method to precisely obtain the transformation between the display and the robot. A path generation algorithm is developed to automatically yield the robot path that can best accommodate the 3D shape of the target model. To further avoid shape distortion due to asynchronization between the display's pose and the image content, we propose a real-time slicing method for arbitrary slicing direction. By organizing the triangular mesh into Octree data structure, the approach can significantly reduce the computational time and improve the performance of real-time rendering. We study the optimal tree level for different ranges of triangle numbers so as to attain competitive computational time. Texture mapping is also implemented to produce colored light painting. We extend our methodologies to computational kinetic photography, which is dual to light painting. Instead of keeping the camera stationary, we move the camera with robot and capture long exposures of a stationary display showing light contours. We transform the display path for light painting to the camera path for kinetic photography. A variety of 3D models are used to verify that the proposed techniques can produce stunning long exposures with high-fidelity volumetric imagery. The techniques have great potential for innovative applications including animation, visible light communication, invisible information visualization and creative art.
We present a semi-automatic method for creating shades and self-shadows in cel animation. Besides producing attractive images, shades and shadows provide important visual cues about depth, shapes, movement and lighting of the scene. In conventional cel animation, shades and shadows are drawn by hand. As opposed to previous approaches, this method does not rely on a complex 3D reconstruction of the scene: its key advantages are simplicity and ease of use. The tool was designed to stay as close as possible to the natural 2D creative environment and therefore provides an intuitive and user-friendly interface. Our system creates shading based on hand-drawn objects or characters, given very limited guidance from the user. The method employs simple yet very efficient algorithms to create shading directly out of drawn strokes. We evaluate our system through a subjective user study and provide qualitative comparison of our method versus existing professional tools and state of the art.
Characters in traditional artwork such as children's books or cartoon animations are typically drawn once, in fixed poses, with little opportunity to change the characters' appearance or re-use them in a different animation. To enable these applications one can fit a consistent parametric deformable model--- a puppet ---to different images of a character, thus establishing consistent segmentation, dense semantic correspondence, and deformation parameters across poses. In this work, we argue that a layered deformable puppet is a natural representation for hand-drawn characters, providing an effective way to deal with the articulation, expressive deformation, and occlusion that are common to this style of artwork. Our main contribution is an automatic pipeline for fitting these models to unlabeled images depicting the same character in various poses. We demonstrate that the output of our pipeline can be used directly for editing and re-targeting animations.
Describing the history of a terrain from a vertical geological cross-section is an important problem in geology, called geological restoration. Designing the sequential evolution of the geometry is usually done manually, involving many trials and errors. In this work, we recast this problem as a storyboarding problem, where the different stages in the restoration are automatically generated as storyboard panels and displayed as geological stories. Our system allows geologists to interactively explore multiple scenarios by selecting plausible geological event sequences and backward simulating them at interactive rate, causing the terrain layers to be progressively un-deposited, un-eroded, un-compacted, un-folded and un-faulted. Storyboard sketches are generated along the way. When a restoration is complete, the storyboard panels can be used for automatically generating a forward animation of the terrain history, enabling quick visualization and validation of hypotheses. As a proof-of-concept, we describe how our system was used by geologists to restore and animate cross-sections in real examples at various spatial and temporal scales and with different levels of complexity, including the Chartreuse region in the French Alps.
This artwork is an artificial life simulation that shows how a society of agents flourishes with the symbiotic interactions between the egotist and altruist extremes. Egotist agents seek and absorb energy. Altruist agents seek other agents, share energy and reproduce. They group into multi-agent organisms that adapt to the energy present in the system.
This work presents enhancements to state-of-the-art adaptive neural style transfer techniques, thereby providing a generalized user interface with creativity tool support for lower-level local control to facilitate the demanding interactive editing on mobile devices. The approaches are implemented in a mobile app that is designed for orchestration of three neural style transfer techniques using iterative, multi-style generative and adaptive neural networks that can be locally controlled by on-screen painting metaphors to perform location-based filtering and direct the composition. Based on first user tests, we conclude with insights, showing different levels of satisfaction for the implemented techniques and user interaction design, pointing out directions for future research.
Stereoscopic 3D (S3D) line drawings were introduced by Sir Charles Wheatstone in 1838. S3D line drawings persist today in various art forms, such as comic books.