Michael Zollhöfer

Lukas Höllein, Aljaž Božič, Michael Zollhöfer et al.

We present 3DGS-LM, a new method that accelerates the reconstruction of 3D Gaussian Splatting (3DGS) by replacing its ADAM optimizer with a tailored Levenberg-Marquardt (LM). Existing methods reduce the optimization time by decreasing the number of Gaussians or by improving the implementation of the differentiable rasterizer. However, they still rely on the ADAM optimizer to fit Gaussian parameters of a scene in thousands of iterations, which can take up to an hour. To this end, we change the optimizer to LM that runs in conjunction with the 3DGS differentiable rasterizer. For efficient GPU parallization, we propose a caching data structure for intermediate gradients that allows us to efficiently calculate Jacobian-vector products in custom CUDA kernels. In every LM iteration, we calculate update directions from multiple image subsets using these kernels and combine them in a weighted mean. Overall, our method is 20% faster than the original 3DGS while obtaining the same reconstruction quality. Our optimization is also agnostic to other methods that acclerate 3DGS, thus enabling even faster speedups compared to vanilla 3DGS.

Fangyin Wei, Rohan Chabra, Lingni Ma et al.

Learning geometry, motion, and appearance priors of object classes is important for the solution of a large variety of computer vision problems. While the majority of approaches has focused on static objects, dynamic objects, especially with controllable articulation, are less explored. We propose a novel approach for learning a representation of the geometry, appearance, and motion of a class of articulated objects given only a set of color images as input. In a self-supervised manner, our novel representation learns shape, appearance, and articulation codes that enable independent control of these semantic dimensions. Our model is trained end-to-end without requiring any articulation annotations. Experiments show that our approach performs well for different joint types, such as revolute and prismatic joints, as well as different combinations of these joints. Compared to state of the art that uses direct 3D supervision and does not output appearance, we recover more faithful geometry and appearance from 2D observations only. In addition, our representation enables a large variety of applications, such as few-shot reconstruction, the generation of novel articulations, and novel view-synthesis.

Linning Xu, Vasu Agrawal, William Laney et al.

We present an end-to-end system for the high-fidelity capture, model reconstruction, and real-time rendering of walkable spaces in virtual reality using neural radiance fields. To this end, we designed and built a custom multi-camera rig to densely capture walkable spaces in high fidelity and with multi-view high dynamic range images in unprecedented quality and density. We extend instant neural graphics primitives with a novel perceptual color space for learning accurate HDR appearance, and an efficient mip-mapping mechanism for level-of-detail rendering with anti-aliasing, while carefully optimizing the trade-off between quality and speed. Our multi-GPU renderer enables high-fidelity volume rendering of our neural radiance field model at the full VR resolution of dual 2K$\times$2K at 36 Hz on our custom demo machine. We demonstrate the quality of our results on our challenging high-fidelity datasets, and compare our method and datasets to existing baselines. We release our dataset on our project website.

Linjie Lyu, Ayush Tewari, Marc Habermann et al.

Inverse rendering, the process of inferring scene properties from images, is a challenging inverse problem. The task is ill-posed, as many different scene configurations can give rise to the same image. Most existing solutions incorporate priors into the inverse-rendering pipeline to encourage plausible solutions, but they do not consider the inherent ambiguities and the multi-modal distribution of possible decompositions. In this work, we propose a novel scheme that integrates a denoising diffusion probabilistic model pre-trained on natural illumination maps into an optimization framework involving a differentiable path tracer. The proposed method allows sampling from combinations of illumination and spatially-varying surface materials that are, both, natural and explain the image observations. We further conduct an extensive comparative study of different priors on illumination used in previous work on inverse rendering. Our method excels in recovering materials and producing highly realistic and diverse environment map samples that faithfully explain the illumination of the input images.

Andreas Kurz, Thomas Neff, Zhaoyang Lv et al.

Novel view synthesis has recently been revolutionized by learning neural radiance fields directly from sparse observations. However, rendering images with this new paradigm is slow due to the fact that an accurate quadrature of the volume rendering equation requires a large number of samples for each ray. Previous work has mainly focused on speeding up the network evaluations that are associated with each sample point, e.g., via caching of radiance values into explicit spatial data structures, but this comes at the expense of model compactness. In this paper, we propose a novel dual-network architecture that takes an orthogonal direction by learning how to best reduce the number of required sample points. To this end, we split our network into a sampling and shading network that are jointly trained. Our training scheme employs fixed sample positions along each ray, and incrementally introduces sparsity throughout training to achieve high quality even at low sample counts. After fine-tuning with the target number of samples, the resulting compact neural representation can be rendered in real-time. Our experiments demonstrate that our approach outperforms concurrent compact neural representations in terms of quality and frame rate and performs on par with highly efficient hybrid representations. Code and supplementary material is available at https://thomasneff.github.io/adanerf.

Wojciech Zielonka, Timur Bagautdinov, Shunsuke Saito et al.

We present Drivable 3D Gaussian Avatars (D3GA), a multi-layered 3D controllable model for human bodies that utilizes 3D Gaussian primitives embedded into tetrahedral cages. The advantage of using cages compared to commonly employed linear blend skinning (LBS) is that primitives like 3D Gaussians are naturally re-oriented and their kernels are stretched via the deformation gradients of the encapsulating tetrahedron. Additional offsets are modeled for the tetrahedron vertices, effectively decoupling the low-dimensional driving poses from the extensive set of primitives to be rendered. This separation is achieved through the localized influence of each tetrahedron on 3D Gaussians, resulting in improved optimization. Using the cage-based deformation model, we introduce a compositional pipeline that decomposes an avatar into layers, such as garments, hands, or faces, improving the modeling of phenomena like garment sliding. These parts can be conditioned on different driving signals, such as keypoints for facial expressions or joint-angle vectors for garments and the body. Our experiments on two multi-view datasets with varied body shapes, clothes, and motions show higher-quality results. They surpass PSNR and SSIM metrics of other SOTA methods using the same data while offering greater flexibility and compactness.

Haithem Turki, Michael Zollhöfer, Christian Richardt et al.

Neural Radiance Fields (NeRFs) can be dramatically accelerated by spatial grid representations. However, they do not explicitly reason about scale and so introduce aliasing artifacts when reconstructing scenes captured at different camera distances. Mip-NeRF and its extensions propose scale-aware renderers that project volumetric frustums rather than point samples but such approaches rely on positional encodings that are not readily compatible with grid methods. We propose a simple modification to grid-based models by training model heads at different spatial grid resolutions. At render time, we simply use coarser grids to render samples that cover larger volumes. Our method can be easily applied to existing accelerated NeRF methods and significantly improves rendering quality (reducing error rates by 20-90% across synthetic and unbounded real-world scenes) while incurring minimal performance overhead (as each model head is quick to evaluate). Compared to Mip-NeRF, we reduce error rates by 20% while training over 60x faster.

Ziqian Bai, Timur Bagautdinov, Javier Romero et al.

Neural fields such as implicit surfaces have recently enabled avatar modeling from raw scans without explicit temporal correspondences. In this work, we exploit autoregressive modeling to further extend this notion to capture dynamic effects, such as soft-tissue deformations. Although autoregressive models are naturally capable of handling dynamics, it is non-trivial to apply them to implicit representations, as explicit state decoding is infeasible due to prohibitive memory requirements. In this work, for the first time, we enable autoregressive modeling of implicit avatars. To reduce the memory bottleneck and efficiently model dynamic implicit surfaces, we introduce the notion of articulated observer points, which relate implicit states to the explicit surface of a parametric human body model. We demonstrate that encoding implicit surfaces as a set of height fields defined on articulated observer points leads to significantly better generalization compared to a latent representation. The experiments show that our approach outperforms the state of the art, achieving plausible dynamic deformations even for unseen motions. https://zqbai-jeremy.github.io/autoavatar

Linjie Lyu, Ayush Tewari, Marc Habermann et al.

Radiance fields are powerful and, hence, popular models for representing the appearance of complex scenes. Yet, constructing them based on image observations gives rise to ambiguities and uncertainties. We propose a versatile approach for learning Gaussian radiance fields with explicit and fine-grained uncertainty estimates that impose only little additional cost compared to uncertainty-agnostic training. Our key observation is that uncertainties can be modeled as a low-dimensional manifold in the space of radiance field parameters that is highly amenable to Monte Carlo sampling. Importantly, our uncertainties are differentiable and, thus, allow for gradient-based optimization of subsequent captures that optimally reduce ambiguities. We demonstrate state-of-the-art performance on next-best-view planning tasks, including high-dimensional illumination planning for optimal radiance field relighting quality.

Stefano Esposito, Anpei Chen, Christian Reiser et al.

High-quality view synthesis relies on volume rendering, splatting, or surface rendering. While surface rendering is typically the fastest, it struggles to accurately model fuzzy geometry like hair. In turn, alpha-blending techniques excel at representing fuzzy materials but require an unbounded number of samples per ray (P1). Further overheads are induced by empty space skipping in volume rendering (P2) and sorting input primitives in splatting (P3). We present a novel representation for real-time view synthesis where the (P1) number of sampling locations is small and bounded, (P2) sampling locations are efficiently found via rasterization, and (P3) rendering is sorting-free. We achieve this by representing objects as semi-transparent multi-layer meshes rendered in a fixed order. First, we model surface layers as signed distance function (SDF) shells with optimal spacing learned during training. Then, we bake them as meshes and fit UV textures. Unlike single-surface methods, our multi-layer representation effectively models fuzzy objects. In contrast to volume and splatting-based methods, our approach enables real-time rendering on low-power laptops and smartphones.

Edgar Tretschk, Ayush Tewari, Vladislav Golyanik et al.

We present Non-Rigid Neural Radiance Fields (NR-NeRF), a reconstruction and novel view synthesis approach for general non-rigid dynamic scenes. Our approach takes RGB images of a dynamic scene as input (e.g., from a monocular video recording), and creates a high-quality space-time geometry and appearance representation. We show that a single handheld consumer-grade camera is sufficient to synthesize sophisticated renderings of a dynamic scene from novel virtual camera views, e.g. a `bullet-time' video effect. NR-NeRF disentangles the dynamic scene into a canonical volume and its deformation. Scene deformation is implemented as ray bending, where straight rays are deformed non-rigidly. We also propose a novel rigidity network to better constrain rigid regions of the scene, leading to more stable results. The ray bending and rigidity network are trained without explicit supervision. Our formulation enables dense correspondence estimation across views and time, and compelling video editing applications such as motion exaggeration. Our code will be open sourced.

Aljaž Božič, Pablo Palafox, Michael Zollhöfer et al.

We introduce a novel, end-to-end learnable, differentiable non-rigid tracker that enables state-of-the-art non-rigid reconstruction by a learned robust optimization. Given two input RGB-D frames of a non-rigidly moving object, we employ a convolutional neural network to predict dense correspondences and their confidences. These correspondences are used as constraints in an as-rigid-as-possible (ARAP) optimization problem. By enabling gradient back-propagation through the weighted non-linear least squares solver, we are able to learn correspondences and confidences in an end-to-end manner such that they are optimal for the task of non-rigid tracking. Under this formulation, correspondence confidences can be learned via self-supervision, informing a learned robust optimization, where outliers and wrong correspondences are automatically down-weighted to enable effective tracking. Compared to state-of-the-art approaches, our algorithm shows improved reconstruction performance, while simultaneously achieving 85 times faster correspondence prediction than comparable deep-learning based methods. We make our code available.

Lukas Höllein, Aljaž Božič, Norman Müller et al.

3D asset generation is getting massive amounts of attention, inspired by the recent success of text-guided 2D content creation. Existing text-to-3D methods use pretrained text-to-image diffusion models in an optimization problem or fine-tune them on synthetic data, which often results in non-photorealistic 3D objects without backgrounds. In this paper, we present a method that leverages pretrained text-to-image models as a prior, and learn to generate multi-view images in a single denoising process from real-world data. Concretely, we propose to integrate 3D volume-rendering and cross-frame-attention layers into each block of the existing U-Net network of the text-to-image model. Moreover, we design an autoregressive generation that renders more 3D-consistent images at any viewpoint. We train our model on real-world datasets of objects and showcase its capabilities to generate instances with a variety of high-quality shapes and textures in authentic surroundings. Compared to the existing methods, the results generated by our method are consistent, and have favorable visual quality (-30% FID, -37% KID).

Haithem Turki, Vasu Agrawal, Samuel Rota Bulò et al.

Neural radiance fields provide state-of-the-art view synthesis quality but tend to be slow to render. One reason is that they make use of volume rendering, thus requiring many samples (and model queries) per ray at render time. Although this representation is flexible and easy to optimize, most real-world objects can be modeled more efficiently with surfaces instead of volumes, requiring far fewer samples per ray. This observation has spurred considerable progress in surface representations such as signed distance functions, but these may struggle to model semi-opaque and thin structures. We propose a method, HybridNeRF, that leverages the strengths of both representations by rendering most objects as surfaces while modeling the (typically) small fraction of challenging regions volumetrically. We evaluate HybridNeRF against the challenging Eyeful Tower dataset along with other commonly used view synthesis datasets. When comparing to state-of-the-art baselines, including recent rasterization-based approaches, we improve error rates by 15-30% while achieving real-time framerates (at least 36 FPS) for virtual-reality resolutions (2Kx2K).

Li Ma, Vasu Agrawal, Haithem Turki et al.

Neural radiance fields have achieved remarkable performance in modeling the appearance of 3D scenes. However, existing approaches still struggle with the view-dependent appearance of glossy surfaces, especially under complex lighting of indoor environments. Unlike existing methods, which typically assume distant lighting like an environment map, we propose a learnable Gaussian directional encoding to better model the view-dependent effects under near-field lighting conditions. Importantly, our new directional encoding captures the spatially-varying nature of near-field lighting and emulates the behavior of prefiltered environment maps. As a result, it enables the efficient evaluation of preconvolved specular color at any 3D location with varying roughness coefficients. We further introduce a data-driven geometry prior that helps alleviate the shape radiance ambiguity in reflection modeling. We show that our Gaussian directional encoding and geometry prior significantly improve the modeling of challenging specular reflections in neural radiance fields, which helps decompose appearance into more physically meaningful components.

Zhaoxi Chen, Gyeongsik Moon, Kaiwen Guo et al.

Existing photorealistic relightable hand models require extensive identity-specific observations in different views, poses, and illuminations, and face challenges in generalizing to natural illuminations and novel identities. To bridge this gap, we present URHand, the first universal relightable hand model that generalizes across viewpoints, poses, illuminations, and identities. Our model allows few-shot personalization using images captured with a mobile phone, and is ready to be photorealistically rendered under novel illuminations. To simplify the personalization process while retaining photorealism, we build a powerful universal relightable prior based on neural relighting from multi-view images of hands captured in a light stage with hundreds of identities. The key challenge is scaling the cross-identity training while maintaining personalized fidelity and sharp details without compromising generalization under natural illuminations. To this end, we propose a spatially varying linear lighting model as the neural renderer that takes physics-inspired shading as input feature. By removing non-linear activations and bias, our specifically designed lighting model explicitly keeps the linearity of light transport. This enables single-stage training from light-stage data while generalizing to real-time rendering under arbitrary continuous illuminations across diverse identities. In addition, we introduce the joint learning of a physically based model and our neural relighting model, which further improves fidelity and generalization. Extensive experiments show that our approach achieves superior performance over existing methods in terms of both quality and generalizability. We also demonstrate quick personalization of URHand from a short phone scan of an unseen identity.

Ethan Weber, Norman Müller, Yash Kant et al.

We present Fillerbuster, a method that completes unknown regions of a 3D scene by utilizing a novel large-scale multi-view latent diffusion transformer. Casual captures are often sparse and miss surrounding content behind objects or above the scene. Existing methods are not suitable for handling this challenge as they focus on making the known pixels look good with sparse-view priors, or on creating the missing sides of objects from just one or two photos. In reality, we often have hundreds of input frames and want to complete areas that are missing and unobserved from the input frames. Additionally, the images often do not have known camera parameters. Our solution is to train a generative model that can consume a large context of input frames while generating unknown target views and recovering image poses when desired. We show results where we complete partial captures on two existing datasets. We also present an uncalibrated scene completion task where our unified model predicts both poses and creates new content. Our model is the first to predict many images and poses together for scene completion.

Chih-Hao Lin, Jia-Bin Huang, Zhengqin Li et al.

Inverse rendering seeks to recover 3D geometry, surface material, and lighting from captured images, enabling advanced applications such as novel-view synthesis, relighting, and virtual object insertion. However, most existing techniques rely on high dynamic range (HDR) images as input, limiting accessibility for general users. In response, we introduce IRIS, an inverse rendering framework that recovers the physically based material, spatially-varying HDR lighting, and camera response functions from multi-view, low-dynamic-range (LDR) images. By eliminating the dependence on HDR input, we make inverse rendering technology more accessible. We evaluate our approach on real-world and synthetic scenes and compare it with state-of-the-art methods. Our results show that IRIS effectively recovers HDR lighting, accurate material, and plausible camera response functions, supporting photorealistic relighting and object insertion.

Ziyan Wang, Timur Bagautdinov, Stephen Lombardi et al.

Photorealistic rendering of dynamic humans is an important ability for telepresence systems, virtual shopping, synthetic data generation, and more. Recently, neural rendering methods, which combine techniques from computer graphics and machine learning, have created high-fidelity models of humans and objects. Some of these methods do not produce results with high-enough fidelity for driveable human models (Neural Volumes) whereas others have extremely long rendering times (NeRF). We propose a novel compositional 3D representation that combines the best of previous methods to produce both higher-resolution and faster results. Our representation bridges the gap between discrete and continuous volumetric representations by combining a coarse 3D-structure-aware grid of animation codes with a continuous learned scene function that maps every position and its corresponding local animation code to its view-dependent emitted radiance and local volume density. Differentiable volume rendering is employed to compute photo-realistic novel views of the human head and upper body as well as to train our novel representation end-to-end using only 2D supervision. In addition, we show that the learned dynamic radiance field can be used to synthesize novel unseen expressions based on a global animation code. Our approach achieves state-of-the-art results for synthesizing novel views of dynamic human heads and the upper body.

Guy Gafni, Justus Thies, Michael Zollhöfer et al.

We present dynamic neural radiance fields for modeling the appearance and dynamics of a human face. Digitally modeling and reconstructing a talking human is a key building-block for a variety of applications. Especially, for telepresence applications in AR or VR, a faithful reproduction of the appearance including novel viewpoints or head-poses is required. In contrast to state-of-the-art approaches that model the geometry and material properties explicitly, or are purely image-based, we introduce an implicit representation of the head based on scene representation networks. To handle the dynamics of the face, we combine our scene representation network with a low-dimensional morphable model which provides explicit control over pose and expressions. We use volumetric rendering to generate images from this hybrid representation and demonstrate that such a dynamic neural scene representation can be learned from monocular input data only, without the need of a specialized capture setup. In our experiments, we show that this learned volumetric representation allows for photo-realistic image generation that surpasses the quality of state-of-the-art video-based reenactment methods.

Aljaž Božič, Pablo Palafox, Michael Zollhöfer et al.

We introduce Neural Deformation Graphs for globally-consistent deformation tracking and 3D reconstruction of non-rigid objects. Specifically, we implicitly model a deformation graph via a deep neural network. This neural deformation graph does not rely on any object-specific structure and, thus, can be applied to general non-rigid deformation tracking. Our method globally optimizes this neural graph on a given sequence of depth camera observations of a non-rigidly moving object. Based on explicit viewpoint consistency as well as inter-frame graph and surface consistency constraints, the underlying network is trained in a self-supervised fashion. We additionally optimize for the geometry of the object with an implicit deformable multi-MLP shape representation. Our approach does not assume sequential input data, thus enabling robust tracking of fast motions or even temporally disconnected recordings. Our experiments demonstrate that our Neural Deformation Graphs outperform state-of-the-art non-rigid reconstruction approaches both qualitatively and quantitatively, with 64% improved reconstruction and 62% improved deformation tracking performance.

Ayush Tewari, Mohamed Elgharib, Mallikarjun B R. et al.

Editing of portrait images is a very popular and important research topic with a large variety of applications. For ease of use, control should be provided via a semantically meaningful parameterization that is akin to computer animation controls. The vast majority of existing techniques do not provide such intuitive and fine-grained control, or only enable coarse editing of a single isolated control parameter. Very recently, high-quality semantically controlled editing has been demonstrated, however only on synthetically created StyleGAN images. We present the first approach for embedding real portrait images in the latent space of StyleGAN, which allows for intuitive editing of the head pose, facial expression, and scene illumination in the image. Semantic editing in parameter space is achieved based on StyleRig, a pretrained neural network that maps the control space of a 3D morphable face model to the latent space of the GAN. We design a novel hierarchical non-linear optimization problem to obtain the embedding. An identity preservation energy term allows spatially coherent edits while maintaining facial integrity. Our approach runs at interactive frame rates and thus allows the user to explore the space of possible edits. We evaluate our approach on a wide set of portrait photos, compare it to the current state of the art, and validate the effectiveness of its components in an ablation study.

Edgar Tretschk, Ayush Tewari, Vladislav Golyanik et al.

Implicit surface representations, such as signed-distance functions, combined with deep learning have led to impressive models which can represent detailed shapes of objects with arbitrary topology. Since a continuous function is learned, the reconstructions can also be extracted at any arbitrary resolution. However, large datasets such as ShapeNet are required to train such models. In this paper, we present a new mid-level patch-based surface representation. At the level of patches, objects across different categories share similarities, which leads to more generalizable models. We then introduce a novel method to learn this patch-based representation in a canonical space, such that it is as object-agnostic as possible. We show that our representation trained on one category of objects from ShapeNet can also well represent detailed shapes from any other category. In addition, it can be trained using much fewer shapes, compared to existing approaches. We show several applications of our new representation, including shape interpolation and partial point cloud completion. Due to explicit control over positions, orientations and scales of patches, our representation is also more controllable compared to object-level representations, which enables us to deform encoded shapes non-rigidly.

Justus Thies, Michael Zollhöfer, Marc Stamminger et al.

We present Face2Face, a novel approach for real-time facial reenactment of a monocular target video sequence (e.g., Youtube video). The source sequence is also a monocular video stream, captured live with a commodity webcam. Our goal is to animate the facial expressions of the target video by a source actor and re-render the manipulated output video in a photo-realistic fashion. To this end, we first address the under-constrained problem of facial identity recovery from monocular video by non-rigid model-based bundling. At run time, we track facial expressions of both source and target video using a dense photometric consistency measure. Reenactment is then achieved by fast and efficient deformation transfer between source and target. The mouth interior that best matches the re-targeted expression is retrieved from the target sequence and warped to produce an accurate fit. Finally, we convincingly re-render the synthesized target face on top of the corresponding video stream such that it seamlessly blends with the real-world illumination. We demonstrate our method in a live setup, where Youtube videos are reenacted in real time.

Ayush Tewari, Ohad Fried, Justus Thies et al.

Efficient rendering of photo-realistic virtual worlds is a long standing effort of computer graphics. Modern graphics techniques have succeeded in synthesizing photo-realistic images from hand-crafted scene representations. However, the automatic generation of shape, materials, lighting, and other aspects of scenes remains a challenging problem that, if solved, would make photo-realistic computer graphics more widely accessible. Concurrently, progress in computer vision and machine learning have given rise to a new approach to image synthesis and editing, namely deep generative models. Neural rendering is a new and rapidly emerging field that combines generative machine learning techniques with physical knowledge from computer graphics, e.g., by the integration of differentiable rendering into network training. With a plethora of applications in computer graphics and vision, neural rendering is poised to become a new area in the graphics community, yet no survey of this emerging field exists. This state-of-the-art report summarizes the recent trends and applications of neural rendering. We focus on approaches that combine classic computer graphics techniques with deep generative models to obtain controllable and photo-realistic outputs. Starting with an overview of the underlying computer graphics and machine learning concepts, we discuss critical aspects of neural rendering approaches. This state-of-the-art report is focused on the many important use cases for the described algorithms such as novel view synthesis, semantic photo manipulation, facial and body reenactment, relighting, free-viewpoint video, and the creation of photo-realistic avatars for virtual and augmented reality telepresence. Finally, we conclude with a discussion of the social implications of such technology and investigate open research problems.

Ayush Tewari, Mohamed Elgharib, Gaurav Bharaj et al.

StyleGAN generates photorealistic portrait images of faces with eyes, teeth, hair and context (neck, shoulders, background), but lacks a rig-like control over semantic face parameters that are interpretable in 3D, such as face pose, expressions, and scene illumination. Three-dimensional morphable face models (3DMMs) on the other hand offer control over the semantic parameters, but lack photorealism when rendered and only model the face interior, not other parts of a portrait image (hair, mouth interior, background). We present the first method to provide a face rig-like control over a pretrained and fixed StyleGAN via a 3DMM. A new rigging network, RigNet is trained between the 3DMM's semantic parameters and StyleGAN's input. The network is trained in a self-supervised manner, without the need for manual annotations. At test time, our method generates portrait images with the photorealism of StyleGAN and provides explicit control over the 3D semantic parameters of the face.

Aljaž Božič, Michael Zollhöfer, Christian Theobalt et al.

Applying data-driven approaches to non-rigid 3D reconstruction has been difficult, which we believe can be attributed to the lack of a large-scale training corpus. Unfortunately, this method fails for important cases such as highly non-rigid deformations. We first address this problem of lack of data by introducing a novel semi-supervised strategy to obtain dense inter-frame correspondences from a sparse set of annotations. This way, we obtain a large dataset of 400 scenes, over 390,000 RGB-D frames, and 5,533 densely aligned frame pairs; in addition, we provide a test set along with several metrics for evaluation. Based on this corpus, we introduce a data-driven non-rigid feature matching approach, which we integrate into an optimization-based reconstruction pipeline. Here, we propose a new neural network that operates on RGB-D frames, while maintaining robustness under large non-rigid deformations and producing accurate predictions. Our approach significantly outperforms existing non-rigid reconstruction methods that do not use learned data terms, as well as learning-based approaches that only use self-supervision.

Hyeongwoo Kim, Mohamed Elgharib, Michael Zollhöfer et al.

Dubbing is a technique for translating video content from one language to another. However, state-of-the-art visual dubbing techniques directly copy facial expressions from source to target actors without considering identity-specific idiosyncrasies such as a unique type of smile. We present a style-preserving visual dubbing approach from single video inputs, which maintains the signature style of target actors when modifying facial expressions, including mouth motions, to match foreign languages. At the heart of our approach is the concept of motion style, in particular for facial expressions, i.e., the person-specific expression change that is yet another essential factor beyond visual accuracy in face editing applications. Our method is based on a recurrent generative adversarial network that captures the spatiotemporal co-activation of facial expressions, and enables generating and modifying the facial expressions of the target actor while preserving their style. We train our model with unsynchronized source and target videos in an unsupervised manner using cycle-consistency and mouth expression losses, and synthesize photorealistic video frames using a layered neural face renderer. Our approach generates temporally coherent results, and handles dynamic backgrounds. Our results show that our dubbing approach maintains the idiosyncratic style of the target actor better than previous approaches, even for widely differing source and target actors.

Vincent Sitzmann, Michael Zollhöfer, Gordon Wetzstein

Unsupervised learning with generative models has the potential of discovering rich representations of 3D scenes. While geometric deep learning has explored 3D-structure-aware representations of scene geometry, these models typically require explicit 3D supervision. Emerging neural scene representations can be trained only with posed 2D images, but existing methods ignore the three-dimensional structure of scenes. We propose Scene Representation Networks (SRNs), a continuous, 3D-structure-aware scene representation that encodes both geometry and appearance. SRNs represent scenes as continuous functions that map world coordinates to a feature representation of local scene properties. By formulating the image formation as a differentiable ray-marching algorithm, SRNs can be trained end-to-end from only 2D images and their camera poses, without access to depth or shape. This formulation naturally generalizes across scenes, learning powerful geometry and appearance priors in the process. We demonstrate the potential of SRNs by evaluating them for novel view synthesis, few-shot reconstruction, joint shape and appearance interpolation, and unsupervised discovery of a non-rigid face model.

Ohad Fried, Ayush Tewari, Michael Zollhöfer et al.

Editing talking-head video to change the speech content or to remove filler words is challenging. We propose a novel method to edit talking-head video based on its transcript to produce a realistic output video in which the dialogue of the speaker has been modified, while maintaining a seamless audio-visual flow (i.e. no jump cuts). Our method automatically annotates an input talking-head video with phonemes, visemes, 3D face pose and geometry, reflectance, expression and scene illumination per frame. To edit a video, the user has to only edit the transcript, and an optimization strategy then chooses segments of the input corpus as base material. The annotated parameters corresponding to the selected segments are seamlessly stitched together and used to produce an intermediate video representation in which the lower half of the face is rendered with a parametric face model. Finally, a recurrent video generation network transforms this representation to a photorealistic video that matches the edited transcript. We demonstrate a large variety of edits, such as the addition, removal, and alteration of words, as well as convincing language translation and full sentence synthesis.

Edgar Tretschk, Ayush Tewari, Michael Zollhöfer et al.

Mesh autoencoders are commonly used for dimensionality reduction, sampling and mesh modeling. We propose a general-purpose DEep MEsh Autoencoder (DEMEA) which adds a novel embedded deformation layer to a graph-convolutional mesh autoencoder. The embedded deformation layer (EDL) is a differentiable deformable geometric proxy which explicitly models point displacements of non-rigid deformations in a lower dimensional space and serves as a local rigidity regularizer. DEMEA decouples the parameterization of the deformation from the final mesh resolution since the deformation is defined over a lower dimensional embedded deformation graph. We perform a large-scale study on four different datasets of deformable objects. Reasoning about the local rigidity of meshes using EDL allows us to achieve higher-quality results for highly deformable objects, compared to directly regressing vertex positions. We demonstrate multiple applications of DEMEA, including non-rigid 3D reconstruction from depth and shading cues, non-rigid surface tracking, as well as the transfer of deformations over different meshes.

Justus Thies, Michael Zollhöfer, Matthias Nießner

The modern computer graphics pipeline can synthesize images at remarkable visual quality; however, it requires well-defined, high-quality 3D content as input. In this work, we explore the use of imperfect 3D content, for instance, obtained from photo-metric reconstructions with noisy and incomplete surface geometry, while still aiming to produce photo-realistic (re-)renderings. To address this challenging problem, we introduce Deferred Neural Rendering, a new paradigm for image synthesis that combines the traditional graphics pipeline with learnable components. Specifically, we propose Neural Textures, which are learned feature maps that are trained as part of the scene capture process. Similar to traditional textures, neural textures are stored as maps on top of 3D mesh proxies; however, the high-dimensional feature maps contain significantly more information, which can be interpreted by our new deferred neural rendering pipeline. Both neural textures and deferred neural renderer are trained end-to-end, enabling us to synthesize photo-realistic images even when the original 3D content was imperfect. In contrast to traditional, black-box 2D generative neural networks, our 3D representation gives us explicit control over the generated output, and allows for a wide range of application domains. For instance, we can synthesize temporally-consistent video re-renderings of recorded 3D scenes as our representation is inherently embedded in 3D space. This way, neural textures can be utilized to coherently re-render or manipulate existing video content in both static and dynamic environments at real-time rates. We show the effectiveness of our approach in several experiments on novel view synthesis, scene editing, and facial reenactment, and compare to state-of-the-art approaches that leverage the standard graphics pipeline as well as conventional generative neural networks.

Michael Zollhöfer

Over the past ten years we have seen a democratization of range sensing technology. While previously range sensors have been highly expensive and only accessible to a few domain experts, such sensors are nowadays ubiquitous and can even be found in the latest generation of mobile devices, e.g., current smartphones. This democratization of range sensing technology was started with the release of the Microsoft Kinect, and since then many different commodity range sensors followed its lead, such as the Primesense Carmine, Asus Xtion Pro, and the Structure Sensor from Occipital. The availability of cheap range sensing technology led to a big leap in research, especially in the context of more powerful static and dynamic reconstruction techniques, starting from 3D scanning applications, such as KinectFusion, to highly accurate face and body tracking approaches. In this chapter, we have a detailed look into the different types of existing range sensors. We discuss the two fundamental types of commodity range sensing techniques in detail, namely passive and active sensing, and we explore the principles these technologies are based on. Our focus is on modern active commodity range sensors based on time-of-flight and structured light. We conclude by discussing the noise characteristics, working ranges, and types of errors made by the different sensing modalities.

Ayush Tewari, Florian Bernard, Pablo Garrido et al.

Monocular image-based 3D reconstruction of faces is a long-standing problem in computer vision. Since image data is a 2D projection of a 3D face, the resulting depth ambiguity makes the problem ill-posed. Most existing methods rely on data-driven priors that are built from limited 3D face scans. In contrast, we propose multi-frame video-based self-supervised training of a deep network that (i) learns a face identity model both in shape and appearance while (ii) jointly learning to reconstruct 3D faces. Our face model is learned using only corpora of in-the-wild video clips collected from the Internet. This virtually endless source of training data enables learning of a highly general 3D face model. In order to achieve this, we propose a novel multi-frame consistency loss that ensures consistent shape and appearance across multiple frames of a subject's face, thus minimizing depth ambiguity. At test time we can use an arbitrary number of frames, so that we can perform both monocular as well as multi-frame reconstruction.

Vincent Sitzmann, Justus Thies, Felix Heide et al.

In this work, we address the lack of 3D understanding of generative neural networks by introducing a persistent 3D feature embedding for view synthesis. To this end, we propose DeepVoxels, a learned representation that encodes the view-dependent appearance of a 3D scene without having to explicitly model its geometry. At its core, our approach is based on a Cartesian 3D grid of persistent embedded features that learn to make use of the underlying 3D scene structure. Our approach combines insights from 3D geometric computer vision with recent advances in learning image-to-image mappings based on adversarial loss functions. DeepVoxels is supervised, without requiring a 3D reconstruction of the scene, using a 2D re-rendering loss and enforces perspective and multi-view geometry in a principled manner. We apply our persistent 3D scene representation to the problem of novel view synthesis demonstrating high-quality results for a variety of challenging scenes.

Justus Thies, Michael Zollhöfer, Christian Theobalt et al.

We propose a learned image-guided rendering technique that combines the benefits of image-based rendering and GAN-based image synthesis. The goal of our method is to generate photo-realistic re-renderings of reconstructed objects for virtual and augmented reality applications (e.g., virtual showrooms, virtual tours \& sightseeing, the digital inspection of historical artifacts). A core component of our work is the handling of view-dependent effects. Specifically, we directly train an object-specific deep neural network to synthesize the view-dependent appearance of an object. As input data we are using an RGB video of the object. This video is used to reconstruct a proxy geometry of the object via multi-view stereo. Based on this 3D proxy, the appearance of a captured view can be warped into a new target view as in classical image-based rendering. This warping assumes diffuse surfaces, in case of view-dependent effects, such as specular highlights, it leads to artifacts. To this end, we propose EffectsNet, a deep neural network that predicts view-dependent effects. Based on these estimations, we are able to convert observed images to diffuse images. These diffuse images can be projected into other views. In the target view, our pipeline reinserts the new view-dependent effects. To composite multiple reprojected images to a final output, we learn a composition network that outputs photo-realistic results. Using this image-guided approach, the network does not have to allocate capacity on ``remembering'' object appearance, instead it learns how to combine the appearance of captured images. We demonstrate the effectiveness of our approach both qualitatively and quantitatively on synthetic as well as on real data.

Justus Thies, Michael Zollhöfer, Christian Theobalt et al.

We propose HeadOn, the first real-time source-to-target reenactment approach for complete human portrait videos that enables transfer of torso and head motion, face expression, and eye gaze. Given a short RGB-D video of the target actor, we automatically construct a personalized geometry proxy that embeds a parametric head, eye, and kinematic torso model. A novel real-time reenactment algorithm employs this proxy to photo-realistically map the captured motion from the source actor to the target actor. On top of the coarse geometric proxy, we propose a video-based rendering technique that composites the modified target portrait video via view- and pose-dependent texturing, and creates photo-realistic imagery of the target actor under novel torso and head poses, facial expressions, and gaze directions. To this end, we propose a robust tracking of the face and torso of the source actor. We extensively evaluate our approach and show significant improvements in enabling much greater flexibility in creating realistic reenacted output videos.

Hyeongwoo Kim, Pablo Garrido, Ayush Tewari et al.

We present a novel approach that enables photo-realistic re-animation of portrait videos using only an input video. In contrast to existing approaches that are restricted to manipulations of facial expressions only, we are the first to transfer the full 3D head position, head rotation, face expression, eye gaze, and eye blinking from a source actor to a portrait video of a target actor. The core of our approach is a generative neural network with a novel space-time architecture. The network takes as input synthetic renderings of a parametric face model, based on which it predicts photo-realistic video frames for a given target actor. The realism in this rendering-to-video transfer is achieved by careful adversarial training, and as a result, we can create modified target videos that mimic the behavior of the synthetically-created input. In order to enable source-to-target video re-animation, we render a synthetic target video with the reconstructed head animation parameters from a source video, and feed it into the trained network -- thus taking full control of the target. With the ability to freely recombine source and target parameters, we are able to demonstrate a large variety of video rewrite applications without explicitly modeling hair, body or background. For instance, we can reenact the full head using interactive user-controlled editing, and realize high-fidelity visual dubbing. To demonstrate the high quality of our output, we conduct an extensive series of experiments and evaluations, where for instance a user study shows that our video edits are hard to detect.

Justus Thies, Michael Zollhöfer, Matthias Nießner

We present IMU2Face, a gesture-driven facial reenactment system. To this end, we combine recent advances in facial motion capture and inertial measurement units (IMUs) to control the facial expressions of a person in a target video based on intuitive hand gestures. IMUs are omnipresent, since modern smart-phones, smart-watches and drones integrate such sensors, e.g., for changing the orientation of the screen content, counting steps, or for flight stabilization. Face tracking and reenactment is based on the state-of-the-art real-time Face2Face facial reenactment system. Instead of transferring facial expressions from a source to a target actor, we employ an IMU to track the hand gestures of a source actor and use its orientation to modify the target actor's expressions.

Ayush Tewari, Michael Zollhöfer, Pablo Garrido et al.

The reconstruction of dense 3D models of face geometry and appearance from a single image is highly challenging and ill-posed. To constrain the problem, many approaches rely on strong priors, such as parametric face models learned from limited 3D scan data. However, prior models restrict generalization of the true diversity in facial geometry, skin reflectance and illumination. To alleviate this problem, we present the first approach that jointly learns 1) a regressor for face shape, expression, reflectance and illumination on the basis of 2) a concurrently learned parametric face model. Our multi-level face model combines the advantage of 3D Morphable Models for regularization with the out-of-space generalization of a learned corrective space. We train end-to-end on in-the-wild images without dense annotations by fusing a convolutional encoder with a differentiable expert-designed renderer and a self-supervised training loss, both defined at multiple detail levels. Our approach compares favorably to the state-of-the-art in terms of reconstruction quality, better generalizes to real world faces, and runs at over 250 Hz.

Weipeng Xu, Avishek Chatterjee, Michael Zollhöfer et al.

We present the first marker-less approach for temporally coherent 3D performance capture of a human with general clothing from monocular video. Our approach reconstructs articulated human skeleton motion as well as medium-scale non-rigid surface deformations in general scenes. Human performance capture is a challenging problem due to the large range of articulation, potentially fast motion, and considerable non-rigid deformations, even from multi-view data. Reconstruction from monocular video alone is drastically more challenging, since strong occlusions and the inherent depth ambiguity lead to a highly ill-posed reconstruction problem. We tackle these challenges by a novel approach that employs sparse 2D and 3D human pose detections from a convolutional neural network using a batch-based pose estimation strategy. Joint recovery of per-batch motion allows to resolve the ambiguities of the monocular reconstruction problem based on a low dimensional trajectory subspace. In addition, we propose refinement of the surface geometry based on fully automatically extracted silhouettes to enable medium-scale non-rigid alignment. We demonstrate state-of-the-art performance capture results that enable exciting applications such as video editing and free viewpoint video, previously infeasible from monocular video. Our qualitative and quantitative evaluation demonstrates that our approach significantly outperforms previous monocular methods in terms of accuracy, robustness and scene complexity that can be handled.

Hyeongwoo Kim, Michael Zollhöfer, Ayush Tewari et al.

We introduce InverseFaceNet, a deep convolutional inverse rendering framework for faces that jointly estimates facial pose, shape, expression, reflectance and illumination from a single input image. By estimating all parameters from just a single image, advanced editing possibilities on a single face image, such as appearance editing and relighting, become feasible in real time. Most previous learning-based face reconstruction approaches do not jointly recover all dimensions, or are severely limited in terms of visual quality. In contrast, we propose to recover high-quality facial pose, shape, expression, reflectance and illumination using a deep neural network that is trained using a large, synthetically created training corpus. Our approach builds on a novel loss function that measures model-space similarity directly in parameter space and significantly improves reconstruction accuracy. We further propose a self-supervised bootstrapping process in the network training loop, which iteratively updates the synthetic training corpus to better reflect the distribution of real-world imagery. We demonstrate that this strategy outperforms completely synthetically trained networks. Finally, we show high-quality reconstructions and compare our approach to several state-of-the-art approaches.

Ayush Tewari, Michael Zollhöfer, Hyeongwoo Kim et al.

In this work we propose a novel model-based deep convolutional autoencoder that addresses the highly challenging problem of reconstructing a 3D human face from a single in-the-wild color image. To this end, we combine a convolutional encoder network with an expert-designed generative model that serves as decoder. The core innovation is our new differentiable parametric decoder that encapsulates image formation analytically based on a generative model. Our decoder takes as input a code vector with exactly defined semantic meaning that encodes detailed face pose, shape, expression, skin reflectance and scene illumination. Due to this new way of combining CNN-based with model-based face reconstruction, the CNN-based encoder learns to extract semantically meaningful parameters from a single monocular input image. For the first time, a CNN encoder and an expert-designed generative model can be trained end-to-end in an unsupervised manner, which renders training on very large (unlabeled) real world data feasible. The obtained reconstructions compare favorably to current state-of-the-art approaches in terms of quality and richness of representation.

Lucas Thies, Michael Zollhöfer, Christian Richardt et al.

We present a novel approach for real-time joint reconstruction of 3D scene motion and geometry from binocular stereo videos. Our approach is based on a novel variational halfway-domain scene flow formulation, which allows us to obtain highly accurate spatiotemporal reconstructions of shape and motion. We solve the underlying optimization problem at real-time frame rates using a novel data-parallel robust non-linear optimization strategy. Fast convergence and large displacement flows are achieved by employing a novel hierarchy that stores delta flows between hierarchy levels. High performance is obtained by the introduction of a coarser warp grid that decouples the number of unknowns from the input resolution of the images. We demonstrate our approach in a live setup that is based on two commodity webcams, as well as on publicly available video data. Our extensive experiments and evaluations show that our approach produces high-quality dense reconstructions of 3D geometry and scene flow at real-time frame rates, and compares favorably to the state of the art.

Srinath Sridhar, Franziska Mueller, Michael Zollhöfer et al.

Real-time simultaneous tracking of hands manipulating and interacting with external objects has many potential applications in augmented reality, tangible computing, and wearable computing. However, due to difficult occlusions, fast motions, and uniform hand appearance, jointly tracking hand and object pose is more challenging than tracking either of the two separately. Many previous approaches resort to complex multi-camera setups to remedy the occlusion problem and often employ expensive segmentation and optimization steps which makes real-time tracking impossible. In this paper, we propose a real-time solution that uses a single commodity RGB-D camera. The core of our approach is a 3D articulated Gaussian mixture alignment strategy tailored to hand-object tracking that allows fast pose optimization. The alignment energy uses novel regularizers to address occlusions and hand-object contacts. For added robustness, we guide the optimization with discriminative part classification of the hand and segmentation of the object. We conducted extensive experiments on several existing datasets and introduce a new annotated hand-object dataset. Quantitative and qualitative results show the key advantages of our method: speed, accuracy, and robustness.

Justus Thies, Michael Zollhöfer, Marc Stamminger et al.

We propose FaceVR, a novel image-based method that enables video teleconferencing in VR based on self-reenactment. State-of-the-art face tracking methods in the VR context are focused on the animation of rigged 3d avatars. While they achieve good tracking performance the results look cartoonish and not real. In contrast to these model-based approaches, FaceVR enables VR teleconferencing using an image-based technique that results in nearly photo-realistic outputs. The key component of FaceVR is a robust algorithm to perform real-time facial motion capture of an actor who is wearing a head-mounted display (HMD), as well as a new data-driven approach for eye tracking from monocular videos. Based on reenactment of a prerecorded stereo video of the person without the HMD, FaceVR incorporates photo-realistic re-rendering in real time, thus allowing artificial modifications of face and eye appearances. For instance, we can alter facial expressions or change gaze directions in the prerecorded target video. In a live setup, we apply these newly-introduced algorithmic components.

Zachary DeVito, Michael Mara, Michael Zollhöfer et al.

Many graphics and vision problems can be expressed as non-linear least squares optimizations of objective functions over visual data, such as images and meshes. The mathematical descriptions of these functions are extremely concise, but their implementation in real code is tedious, especially when optimized for real-time performance on modern GPUs in interactive applications. In this work, we propose a new language, Opt (available under http://optlang.org), for writing these objective functions over image- or graph-structured unknowns concisely and at a high level. Our compiler automatically transforms these specifications into state-of-the-art GPU solvers based on Gauss-Newton or Levenberg-Marquardt methods. Opt can generate different variations of the solver, so users can easily explore tradeoffs in numerical precision, matrix-free methods, and solver approaches. In our results, we implement a variety of real-world graphics and vision applications. Their energy functions are expressible in tens of lines of code, and produce highly-optimized GPU solver implementations. These solver have performance competitive with the best published hand-tuned, application-specific GPU solvers, and orders of magnitude beyond a general-purpose auto-generated solver.

Angela Dai, Matthias Nießner, Michael Zollhöfer et al.

Real-time, high-quality, 3D scanning of large-scale scenes is key to mixed reality and robotic applications. However, scalability brings challenges of drift in pose estimation, introducing significant errors in the accumulated model. Approaches often require hours of offline processing to globally correct model errors. Recent online methods demonstrate compelling results, but suffer from: (1) needing minutes to perform online correction preventing true real-time use; (2) brittle frame-to-frame (or frame-to-model) pose estimation resulting in many tracking failures; or (3) supporting only unstructured point-based representations, which limit scan quality and applicability. We systematically address these issues with a novel, real-time, end-to-end reconstruction framework. At its core is a robust pose estimation strategy, optimizing per frame for a global set of camera poses by considering the complete history of RGB-D input with an efficient hierarchical approach. We remove the heavy reliance on temporal tracking, and continually localize to the globally optimized frames instead. We contribute a parallelizable optimization framework, which employs correspondences based on sparse features and dense geometric and photometric matching. Our approach estimates globally optimized (i.e., bundle adjusted) poses in real-time, supports robust tracking with recovery from gross tracking failures (i.e., relocalization), and re-estimates the 3D model in real-time to ensure global consistency; all within a single framework. Our approach outperforms state-of-the-art online systems with quality on par to offline methods, but with unprecedented speed and scan completeness. Our framework leads to a comprehensive online scanning solution for large indoor environments, enabling ease of use and high-quality results.

Matthias Innmann, Michael Zollhöfer, Matthias Nießner et al.

We present a novel approach for the reconstruction of dynamic geometric shapes using a single hand-held consumer-grade RGB-D sensor at real-time rates. Our method does not require a pre-defined shape template to start with and builds up the scene model from scratch during the scanning process. Geometry and motion are parameterized in a unified manner by a volumetric representation that encodes a distance field of the surface geometry as well as the non-rigid space deformation. Motion tracking is based on a set of extracted sparse color features in combination with a dense depth-based constraint formulation. This enables accurate tracking and drastically reduces drift inherent to standard model-to-depth alignment. We cast finding the optimal deformation of space as a non-linear regularized variational optimization problem by enforcing local smoothness and proximity to the input constraints. The problem is tackled in real-time at the camera's capture rate using a data-parallel flip-flop optimization strategy. Our results demonstrate robust tracking even for fast motion and scenes that lack geometric features.