Search

We introduce a neural relighting algorithm for captured indoors scenes, that allows interactive free-viewpoint navigation. Our method allows illumination to be changed synthetically, while coherently rendering cast shadows and complex glossy materials. We start with multiple images of the scene and a three-dimensional mesh obtained by multi-view stereo (MVS) reconstruction. We assume that lighting is well explained as the sum of a view-independent diffuse component and a view-dependent glossy term concentrated around the mirror reflection direction. We design a convolutional network around input feature maps that facilitate learning of an implicit representation of scene materials and illumination, enabling both relighting and free-viewpoint navigation. We generate these input maps by exploiting the best elements of both image-based and physically based rendering. We sample the input views to estimate diffuse scene irradiance, and compute the new illumination caused by user-specified light sources using path tracing. To facilitate the network's understanding of materials and synthesize plausible glossy reflections, we reproject the views and compute mirror images . We train the network on a synthetic dataset where each scene is also reconstructed with MVS. We show results of our algorithm relighting real indoor scenes and performing free-viewpoint navigation with complex and realistic glossy reflections, which so far remained out of reach for view-synthesis techniques.

International audience; Rendering glinty details from specular microstructure enhances the level of realism, but previous methods require heavy storage for the high-resolution height field or normal map and associated acceleration structures. In this article, we aim at dynamically generating theoretically infinite microstructure, preventing obvious tiling artifacts, while achieving constant storage cost. Unlike traditional texture synthesis, our method supports arbitrary point and range queries, and is essentially generating the microstructure implicitly. Our method fits the widely used microfacet rendering framework with multiple importance sampling (MIS), replacing the commonly used microfacet normal distribution functions (NDFs) like ground glass distribution (GGX) by a detailed local solution, with a small amount of runtime performance overhead.

There has recently been great interest in neural rendering methods. Some approaches use 3D geometry reconstructed with Multi-View Stereo (MVS) but cannot recover from the errors of this process, while others directly learn a volumetric neural representation, but suffer from expensive training and inference. We introduce a general approach that is initialized with MVS, but allows further optimization of scene properties in the space of input views, including depth and reprojected features, resulting in improved novel-view synthesis. A key element of our approach is our new differentiable point-based pipeline, based on bi-directional Elliptical Weighted Average splatting, a probabilistic depth test and effective camera selection. We use these elements together in our neural renderer, that outperforms all previous methods both in quality and speed in almost all scenes we tested. Our pipeline can be applied to multi-view harmonization and stylization in addition to novel-view synthesis., https://repo-sam.inria.fr/fungraph/differentiable-multi-view/

Creative live streams, where artists or designers demonstrate their creative process, have emerged as a unique and popular genre of live streams due to the real-time interactivity they afford. However, streamer-viewer interactions on most live streaming platforms only enable users to utilize text and emojis to communicate, which limits what viewers can convey and share in real time. To investigate the design space of potential visual and non-textual modalities within creative live streams, we first analyzed existing Twitch extensions and conducted a formative study with streamers who share creative activities to uncover key challenges that these streamers face. We then designed and implemented a prototype system, StreamSketch, which enables viewers and streamers to interact during live streams using multiple modalities, including freeform sketches and text. The prototype was evaluated by two professional artist streamers and their viewers during six streaming sessions. Overall, streamers and viewers found that StreamSketch provided increased engagement and new affordances compared to the traditional text-only modality, and highlighted how efficiency, moderation, and tool integration were continued challenges.

International audience; Image-Based Rendering allows users to easily capture a scene using a single camera and then navigate freely with realistic results. However, the resulting renderings are completely static, and dynamic effects – such as fire, waterfalls or small waves – cannot be reproduced. We tackle the challenging problem of enabling free-viewpoint navigation including such stationary dynamic effects, but still maintaining the simplicity of casual capture. Using a single camera – instead of previous complex synchronized multi-camera setups – means that we have unsynchronized videos of the dynamic effect from multiple views, making it hard to blend them when synthesizing novel views. We present a solution that allows smooth free-viewpoint video-based rendering (VBR) of such scenes using temporal Laplacian pyramid decomposition video, enabling spatio-temporal blending. For effects such as fire and waterfalls, that are semi-transparent and occupy 3D space, we first estimate their spatial volume. This allows us to create per-video geometries and alpha-matte videos that we can blend using our frequency-dependent method. We also extend Laplacian blending to the temporal dimension to remove additional temporal seams. We show results on scenes containing fire, waterfalls or rippling waves at the seaside, bringing these scenes to life.

International audience; We propose a benchmark of state-of-the-art sound event detection systems (SED). We designed synthetic evaluation sets to focus on specific sound event detection challenges. We analyze the performance of the submissions to DCASE 2021 task 4 depending on time related modifications (time position of an event and length of clips) and we study the impact of non-target sound events and reverberation. We show that the localization in time of sound events is still a problem for SED systems. We also show that reverberation and non-target sound events are severely degrading the performance of the SED systems. In the latter case, sound separation seems like a promising solution.

Keyframe-based sculpting provides unprecedented freedom to author animated organic models, which can be difficult to create with other methods such as simulation, scripting, and rigging. However, sculpting animated objects can require significant artistic skill and manual labor, even more so than sculpting static 3D shapes or drawing 2D animations, which are already quite challenging. We present a keyframe-based animated sculpting system with the capability to autocomplete user editing under a simple and intuitive brushing interface. Similar to current desktop sculpting and VR brushing tools, users can brush surface details and volume structures. Meanwhile, our system analyzes their workflows and predicts what they might do in the future, both spatially and temporally. Users can accept or ignore these suggestions and thus maintain full control. We propose the first interactive suggestive keyframe sculpting system, specifically for spatio-temporal repetitive tasks, including low-level spatial details and high-level brushing structures across multiple frames. Our key ideas include a deformation-based optimization framework to analyze recorded workflows and synthesize predictions, and a semi-causal global similarity measurement to support flexible brushing stroke sequences and complex shape changes. Our system supports a variety of shape and motion styles, including those difficult to achieve via existing animation systems, such as topological changes that cannot be accomplished via simple rig-based deformations and stylized physically-implausible motions that cannot be simulated. We evaluate our system via a pilot user study that demonstrates the effectiveness of our system.

International audience; Aggregate elements are ubiquitous in natural and man-made objects. Interactively authoring these elements with varying anisotropy and deformability can require high artistic skills and manual labor. To reduce input workload and enhance output quality, we present an autocomplete system that can help users distribute and align such elements over different domains. Through a brushing interface, users can place and mix a few elements, and let our system automatically populate more elements for the remaining output. Furthermore, aggregate elements often require proper direction/scalar fields for proper arrangements, but fully specifying such fields across entire domains can be difficult or inconvenient for ordinary users. To address this usability challenge, we formulate "element fields" that can smoothly orient all the elements based on partial user specifications without requiring full input fields in any step.We validate our prototype system with a pilot user study and show applications in design, collage, and modeling.

International audience; We present a comparative analysis of the performance of state-of-the-art sound event detection systems. In particular, we study the robustness of the systems to noise and signal degradation, which is known to impact model generalization. Our analysis is based on the results of task 4 of the DCASE 2019 challenge, where submitted systems were evaluated on, in addition to real-world recordings, a series of synthetic soundscapes that allow us to carefully control for different soundscape characteristics. Our results show that while overall systems exhibit significant improvements compared to previous work, they still suffer from biases that could prevent them from generalizing to real-world scenarios.

International audience; We introduce the Free Universal Sound Separation (FUSS) dataset, a new corpus for experiments in separating mixtures of an unknown number of sounds from an open domain of sound types. The dataset consists of 23 hours of single-source audio data drawn from 357 classes, which are used to create mixtures of one to four sources. To simulate reverberation, an acoustic room simulator is used to generate impulse responses of box shaped rooms with frequency-dependent reflective walls. Additional open-source data augmentation tools are also provided to produce new mixtures with different combinations of sources and room simulations. Finally, we introduce an open-source baseline separation model, based on an improved time-domain convolutional network (TDCN++), that can separate a variable number of sources in a mixture. This model achieves 9.8 dB of scale-invariant signal-to-noise ratio improvement (SI-SNRi) on mixtures with two to four sources, while reconstructing single-source inputs with 35.5 dB absolute SI-SNR. We hope this dataset will lower the barrier to new research and allow for fast iteration and application of novel techniques from other machine learning domains to the sound separation challenge.

Augmented Reality (AR) superimposes digital content on top of the real world, to enhance it and provide a new generation of media experiences. To provide a realistic AR experience, objects in the scene should be delivered with both high photorealism and low latency. Current AR experiences are mostly delivered with a download-and-play strategy, where the whole scene is considered a monolithic entity for delivery. This approach results in high start-up latencies and therefore a poor user experience. A similar problem in the video domain has already been tackled with the HTTP Adaptive Streaming (HAS) principle, where the video is split into segments, and a rate adaptation heuristic dynamically adapts the video quality based on the available network resources. In this paper, we apply the adaptive streaming principle from the video to the AR domain, and propose a streaming framework for AR applications. In our proposed framework, the AR objects are available at different Levels-Of-Detail (LODs) and can be streamed independently from each other. An LOD adaptation heuristic is in charge of dynamically deciding what object should be fetched from the server and at what LOD level. Our proposed heuristic prioritizes content that is more likely to be viewed by the user and selects the best LOD to maximize the object's perceived visual quality. Moreover, the adaptation takes into account the available bandwidth resources to ensure a timely delivery of the AR objects. Experiments carried out over the Internet using an AR streaming prototype developed on an iOS device allow us to show the gains brought by the proposed framework. Particularly, our approach can decrease start-up latency up to 90% with respect to a download-and-play baseline, and decrease the amount of data needed to deliver the AR experience up to 79%, without sacrificing on the visual quality of the AR objects.

International audience; While hair is an essential component of virtual humans, it is also one of the most challenging digital assets to create. Existing automatic techniques lack the generality and flexibility to create rich hair variations, while manual authoring interfaces often require considerable artistic skills and efforts, especially for intricate 3D hair structures that can be difficult to navigate. We propose an interactive hair modeling system that can help create complex hairstyles in minutes or hours that would otherwise take much longer with existing tools. Modelers, including novice users, can focus on the overall hairstyles and local hair deformations, as our system intelligently suggests the desired hair parts. Our method combines the flexibility of manual authoring and the convenience of data-driven automation. Since hair contains intricate 3D structures such as buns, knots, and strands, they are inherently challenging to create using traditional 2D interfaces. Our system provides a new 3D hair author-ing interface for immersive interaction in virtual reality (VR). Users can draw high-level guide strips, from which our system predicts the most plausible hairstyles via a deep neural network trained from a professionally curated dataset. Each hairstyle in our dataset is composed of multiple variations, serving as blend-shapes to fit the user drawings via global blending and local deformation. The fitted hair models are visualized as interactive suggestions that the user can select, modify, or ignore. We conducted a user study to confirm that our system can significantly reduce manual labor while improve the output quality for modeling a variety of head and facial hairstyles that are challenging to create via existing techniques.

International audience; This paper presents Task 4 of the Detection and Classification of Acoustic Scenes and Events (DCASE) 2019 challenge and provides a first analysis of the challenge results. The task is a followup to Task 4 of DCASE 2018, and involves training systems for large-scale detection of sound events using a combination of weakly labeled data, i.e. training labels without time boundaries, and strongly-labeled synthesized data. The paper introduces Domestic Environment Sound Event Detection (DESED) dataset mixing a part of last year dataset and an additional synthetic, strongly labeled, dataset provided this year that we’ll describe more in detail. We also report the performance of the submitted systems on the official evaluation (test) and development sets as well as several additional datasets. The best systems from this year outperform last year’s winning system by about 10% points in terms of F-measure.

International audience; This tutorial describes the geometry and algorithms for generating line drawings from 3D models, focusing on occluding contours. The geometry of occluding contours on meshes and on smooth surfaces is described in detail, together with algorithms for extracting contours, computing their visibility, and creating stylized renderings and animations. Exact methods and hardware-accelerated fast methods are both described, and the trade-offs between different methods are discussed. The tutorial brings together and organizes material that, at present, is scattered throughout the literature. It also includes some novel explanations, and implementation tips. A thorough survey of the field of non-photorealistic 3D rendering is also included, covering other kinds of line drawings and artistic shading.

International audience; We present a system that augments live presentation videos with interactive graphics to create a powerful and expressive storytelling environment.Using our system, the presenter interacts with the graphical elements in real-time with gestures and postures, thus leveraging our innate, everyday skills to enhance our communication capabilities with the audience.However, crafting such an interactive and expressive performance typically requires programming, or highly-specialized tools tailored for experts.Our core contribution is a flexible, direct manipulation UI which enables amateurs and experts to craft such presentations beforehand by mapping a variety of body movements to a wide range of graphical manipulations.By simplifying the mapping between gestures, postures, and their corresponding output effects, our UI enables users to craft customized, rich interactions with the graphical elements.Our user study demonstrates the potential usage and unique affordance of this mixed-reality medium for storytelling and presentation across a range of application domains.

We propose the first learning-based algorithm that can relight images in a plausible and controllable manner given multiple views of an outdoor scene. In particular, we introduce a geometry-aware neural network that utilizes multiple geometry cues (normal maps, specular direction, etc.) and source and target shadow masks computed from a noisy proxy geometry obtained by multi-view stereo. Our model is a three-stage pipeline: two subnetworks refine the source and target shadow masks, and a third performs the final relighting. Furthermore, we introduce a novel representation for the shadow masks, which we call RGB shadow images. They reproject the colors from all views into the shadowed pixels and enable our network to cope with inacuraccies in the proxy and the non-locality of the shadow casting interactions. Acquiring large-scale multi-view relighting datasets for real scenes is challenging, so we train our network on photorealistic synthetic data. At train time, we also compute a noisy stereo-based geometric proxy, this time from the synthetic renderings. This allows us to bridge the gap between the real and synthetic domains. Our model generalizes well to real scenes. It can alter the illumination of drone footage, image-based renderings, textured mesh reconstructions, and even internet photo collections.

Localizing moments in a longer video via natural language queries is a new, challenging task at the intersection of language and video understanding. Though moment localization with natural language is similar to other language and vision tasks like natural language object retrieval in images, moment localization offers an interesting opportunity to model temporal dependencies and reasoning in text. We propose a new model that explicitly reasons about different temporal segments in a video, and shows that temporal context is important for localizing phrases which include temporal language. To benchmark whether our model, and other recent video localization models, can effectively reason about temporal language, we collect the novel TEMPOral reasoning in video and language (TEMPO) dataset. Our dataset consists of two parts: a dataset with real videos and template sentences (TEMPO - Template Language) which allows for controlled studies on temporal language, and a human language dataset which consists of temporal sentences annotated by humans (TEMPO - Human Language)., EMNLP 2018

Aggregate elements following certain directions have a variety\ud of applications in graphics, design, and visualization. However,\ud authoring oriented elements in various output domains, especially\ud in 3D, remains challenging. We propose a novel brushing system to\ud facilitate interactive authoring of aggregate elements with diverse\ud properties over given output domains via an element synthesis\ud approach. To increase output quality and reduce input workload,\ud we further propose element fields that can automatically orient\ud the entire elements in better alignments over the output domains\ud according to partially user-specified strokes. The proposed system\ud can effectively synthesize distinct types of elements within various\ud output domains in higher quality and efficiency and offer more user\ud friendliness than existing practices. Our method can be applied to\ud practical applications such as graphic design, artistic collage, and\ud aggregate modeling

This paper presents an end-to-end convolutional neural network (CNN) for 2D-3D exemplar detection. We demonstrate that the ability to adapt the features of natural images to better align with those of CAD rendered views is critical to the success of our technique. We show that the adaptation can be learned by compositing rendered views of textured object models on natural images. Our approach can be naturally incorporated into a CNN detection pipeline and extends the accuracy and speed benefits from recent advances in deep learning to 2D-3D exemplar detection. We applied our method to two tasks: instance detection, where we evaluated on the IKEA dataset, and object category detection, where we out-perform Aubry et al. for "chair" detection on a subset of the Pascal VOC dataset., To appear in CVPR 2016

Virtual-reality provides an immersive environment but can induce cybersickness due to the discrepancy between visual and vestibular cues. To avoid this problem, the movement of the virtual camera needs to match the motion of the user in the real world. Unfortunately, this is usually difficult due to the mismatch between the size of the virtual environments and the space available to the users in the physical domain. The resulting constraints on the camera movement significantly hamper the adoption of virtual-reality headsets in many scenarios and make the design of the virtual environments very challenging. In this work, we study how the characteristics of the virtual camera movement (e.g., translational acceleration and rotational velocity) and the composition of the virtual environment (e.g., scene depth) contribute to perceived discomfort. Based on the results from our user experiments, we devise a computational model for predicting the magnitude of the discomfort for a given scene and camera trajectory. We further apply our model to a new path planning method which optimizes the input motion trajectory to reduce perceptual sickness. We evaluate the effectiveness of our method in improving perceptual comfort in a series of user studies targeting different applications. The results indicate that our method can reduce the perceived discomfort while maintaining the fidelity of the original navigation, and perform better than simpler alternatives.