Search

We introduce context-aware translation, a novel method that combines the benefits of inpainting and image-to-image translation, respecting simultaneously the original input and contextual relevance -- where existing methods fall short. By doing so, our method opens new avenues for the controllable use of AI within artistic creation, from animation to digital art. As an use case, we apply our method to redraw any hand-drawn animated character eyes based on any design specifications - eyes serve as a focal point that captures viewer attention and conveys a range of emotions, however, the labor-intensive nature of traditional animation often leads to compromises in the complexity and consistency of eye design. Furthermore, we remove the need for production data for training and introduce a new character recognition method that surpasses existing work by not requiring fine-tuning to specific productions. This proposed use case could help maintain consistency throughout production and unlock bolder and more detailed design choices without the production cost drawbacks. A user study shows context-aware translation is preferred over existing work 95.16% of the time.

Fully automatic semantic segmentation of highly specific semantic classes and complex shapes may not meet the accuracy standards demanded by scientists. In such cases, human-centered AI solutions, able to assist operators while preserving human control over complex tasks, are a good trade-off to speed up image labeling while maintaining high accuracy levels. TagLab is an open-source AI-assisted software for annotating large orthoimages which takes advantage of different degrees of automation; it speeds up image annotation from scratch through assisted tools, creates custom fully automatic semantic segmentation models, and, finally, allows the quick edits of automatic predictions. Since the orthoimages analysis applies to several scientific disciplines, TagLab has been designed with a flexible labeling pipeline. We report our results in two different scenarios, marine ecology, and architectural heritage., Comment: Accepted at Human Centered AI workshop at NeurIPS 2021, https://sites.google.com/view/hcai-human-centered-ai-neurips/home

We are witnessing a proliferation of textured 3D models captured from the real world with automatic photo-reconstruction tools. Digital 3D models of this class come with a unique set of characteristics and defects -- especially concerning their parametrization -- setting them starkly apart from 3D models originating from other, more traditional, sources. We study this class of 3D models by collecting a significant number of representatives and quantitatively evaluating their quality according to several metrics. These include a new invariant metric we design to assess the fragmentation of the UV map, one of the main weaknesses hindering the usability of these models. Our results back the widely shared notion that such models are not fit for direct use in downstream applications (such as videogames), and require challenging processing steps. Regrettably, existing automatic geometry processing tools are not always up to the task: for example, we verify that available tools for UV optimization often fail due mesh inconsistencies, geometric and topological noise, excessive resolution, or other factors; moreover, even when an output is produced, it is rarely a significant improvement over the input (according to the aforementioned measures). Therefore, we argue that further advancements are required specifically targeted at this class of models. Towards this goal, we share the models we collected in the form of a new public repository, Real-World Textured Things (RWTT), a benchmark to systematic field-test and compare algorithms. RWTT consists of 568 carefully selected textured 3D models representative of all the main modern off-the-shelf photo-reconstruction tools. The repository is available at http://texturedmesh.isti.cnr.it/ and is browsable by metadata collected during experiments, and comes with a tool, TexMetro, providing the same set of measures for generic UV mapped datasets.

We present a new fully automatic block-decomposition hexahedral meshing algorithm capable of producing high quality meshes that strictly preserve feature curve networks on the input surface and align with an input surface cross-field. We produce all-hex meshes on the vast majority of inputs, and introduce localized non-hex elements only when the surface feature network necessitates those. The input to our framework is a closed surface with a collection of geometric or user-demarcated feature curves and a feature-aligned surface cross-field. Its output is a compact set of blocks whose edges interpolate these features and are loosely aligned with this cross-field. We obtain this block decomposition by cutting the input model using a collection of simple cutting surfaces bounded by closed surface loops. The set of cutting loops spans the input feature curves, ensuring feature preservation, and is obtained using a field-space sampling process. The computed loops are uniformly distributed across the surface, cross orthogonally, and are loosely aligned with the cross-field directions, inducing the desired block decomposition. We validate our method by applying it to a large range of complex inputs and comparing our results to those produced by state-of-the-art alternatives. Contrary to prior approaches, our framework consistently produces high-quality field aligned meshes while strictly preserving geometric or user-specified surface features.

We propose ReviewerNet, an online, interactive visualization system aimed to improve the reviewer selection process in the academic domain. Given a paper submitted for publication, we assume that good candidate reviewers can be chosen among the authors of a small set of relevant and pertinent papers; ReviewerNet supports the construction of such set of papers, by visualizing and exploring a literature citation network. Then, the system helps to select reviewers that are both well distributed in the scientific community and that do not have any conflict-of-interest, by visualising the careers and co-authorship relations of candidate reviewers. The system is publicly available, and it has been evaluated by a set of experienced researchers in the field of Computer Graphics., Comment: 11 pages, 5 figures

We present a method for turning a flash selfie taken with a smartphone into a photograph as if it was taken in a studio setting with uniform lighting. Our method uses a convolutional neural network trained on a set of pairs of photographs acquired in an ad-hoc acquisition campaign. Each pair consists of one photograph of a subject's face taken with the camera flash enabled and another one of the same subject in the same pose illuminated using a photographic studio-lighting setup. We show how our method can amend defects introduced by a close-up camera flash, such as specular highlights, shadows, skin shine, and flattened images.

We introduce HexaLab: a WebGL application for real time visualization, exploration and assessment of hexahedral meshes. HexaLab can be used by simply opening www.hexalab.net. Our visualization tool targets both users and scholars. Practitioners who employ hexmeshes for Finite Element Analysis, can readily check mesh quality and assess its usability for simulation. Researchers involved in mesh generation may use HexaLab to perform a detailed analysis of the mesh structure, isolating weak points and testing new solutions to improve on the state of the art and generate high quality images. To this end, we support a wide variety of visualization and volume inspection tools. Our system offers also immediate access to a repository containing all the publicly available meshes produced with the most recent techniques for hexmesh generation. We believe HexaLab, providing a common tool for visualizing, assessing and distributing results, will push forward the recent strive for replicability in our scientific community.

We introduce a framework for the generation of grid-shell structures that is based on Voronoi diagrams and allows us to design tessellations that achieve excellent static performances. We start from an analysis of stress on the input surface and we use the resulting tensor field to induce an anisotropic non-Euclidean metric over it. Then we compute a Centroidal Voronoi Tessellation under the same metric. The resulting mesh is hex-dominant and made of cells with a variable density, which depends on the amount of stress, and anisotropic shape, which depends on the direction of maximum stress. This mesh is further optimized taking into account symmetry and regularity of cells to improve aesthetics. We demonstrate that our grid-shells achieve better static performances with respect to quad-based grid shells, while offering an innovative and aesthetically pleasing look., Comment: 10 pages