Monocular facial performance capture in-the-wild is challenging due to varied capture conditions, face shapes, and expressions. Most current methods rely on linear 3D Morphable Models, which represent facial expressions independently of identity at the vertex displacement level. We propose SEREP (Semantic Expression Representation), a model that disentangles expression from identity at the semantic level. We start by learning an expression representation from high-quality 3D data of unpaired facial expressions. Then, we train a model to predict expression from monocular images relying on a novel semi-supervised scheme using low quality synthetic data. In addition, we introduce MultiREX, a benchmark addressing the lack of evaluation resources for the expression capture task. Our experiments show that SEREP outperforms state-of-the-art methods, capturing challenging expressions and transferring them to new identities.

Reconstructing an avatar from a portrait image has many applications in multimedia, but remains a challenging research problem. Extracting reflectance maps and geom- etry from one image is ill-posed: recovering geometry is a one-to-many mapping problem and reflectance and light are difficult to disentangle. Accurate geometry and reflectance can be captured under the controlled conditions of a light stage, but it is costly to acquire large datasets in this fash- ion. Moreover, training solely with this type of data leads to poor generalization with in-the-wild images. This moti- vates the introduction of MoSAR, a method for 3D avatar generation from monocular images. We propose a semi- supervised training scheme that improves generalization by learning from both light stage and in-the-wild datasets. This is achieved using a novel differentiable shading formulation. We show that our approach effectively disentangles the intrinsic face parameters, producing relightable avatars. As a result, MoSAR11Project page: https://ubisoft-laforge.github.io/character/mosar estimates a richer set of skin reflectance maps and generates more realistic avatars than existing state-of-the-art methods. We also release a new dataset, that provides intrinsic face attributes (diffuse, specular, am- bient occlusion and translucency maps) for 10k subjects.

We present a novel face reconstruction method capable of reconstructing detailed face geometry, spatially varying face reflectance from a single monocular image. We build our work upon the recent advances of DNN-based auto-encoders with differentiable ray tracing image formation, trained in self-supervised manner. While providing the advantage of learning-based approaches and real-time reconstruction, the latter methods lacked fidelity. In this work, we achieve, for the first time, high fidelity face reconstruction using self-supervised learning only. Our novel coarse-to-fine deep architecture allows us to solve the challenging problem of decoupling face reflectance from geometry using a single image, at high computational speed. Compared to state-of-the-art methods, our method achieves more visually appealing reconstruction.

Robust face reconstruction from monocular image in general lighting conditions is challenging. Methods combining deep neural network encoders with differentiable rendering have opened up the path for very fast monocular reconstruction of geometry, lighting and reflectance. They can also be trained in self-supervised manner for increased robustness and better generalization. However, their differentiable rasterization-based image formation models, as well as underlying scene parameterization, limit them to Lambertian face reflectance and to poor shape details. More recently, ray tracing was introduced for monocular face reconstruction within a classic optimization-based framework and enables state-of-the art results. However, optimization-based approaches are inherently slow and lack robustness. In this paper, we build our work on the afore-mentioned approaches and propose a new method that greatly improves reconstruction quality and robustness in general scenes. We achieve this by combining a CNN encoder with a differentiable ray tracer, which enables us to base the reconstruction on much more advanced personalized diffuse and specular albedos, a more sophisticated illumination model and a plausible representation of self-shadows. This enables to take a big leap forward in reconstruction quality of shape, appearance and lighting even in scenes with difficult illumination. With consistent face attributes reconstruction, our method leads to practical applications such as relighting and self-shadows removal. Compared to state-of-the-art methods, our results show improved accuracy and validity of the approach.

Abstract We present a differentiable ray‐tracing based novel face reconstruction approach where scene attributes – 3D geometry, reflectance (diffuse, specular and roughness), pose, camera parameters, and scene illumination – are estimated from unconstrained monocular images. The proposed method models scene illumination via a novel, parameterized virtual light stage, which in‐conjunction with differentiable ray‐tracing, introduces a coarse‐to‐fine optimization formulation for face reconstruction. Our method can not only handle unconstrained illumination and self‐shadows conditions, but also estimates diffuse and specular albedos. To estimate the face attributes consistently and with practical semantics, a two‐stage optimization strategy systematically uses a subset of parametric attributes, where subsequent attribute estimations factor those previously estimated. For example, self‐shadows estimated during the first stage, later prevent its baking into the personalized diffuse and specular albedos in the second stage. We show the efficacy of our approach in several real‐world scenarios, where face attributes can be estimated even under extreme illumination conditions. Ablation studies, analyses and comparisons against several recent state‐of‐the‐art methods show improved accuracy and versatility of our approach. With consistent face attributes reconstruction, our method leads to several style – illumination, albedo, self‐shadow – edit and transfer applications, as discussed in the paper.

Photorealistic editing of head portraits is a challenging task as humans are very sensitive to inconsistencies in faces. We present an approach for high-quality intuitive editing of the camera viewpoint and scene illumination (parameterised with an environment map) in a portrait image. This requires our method to capture and control the full reflectance field of the person in the image. Most editing approaches rely on supervised learning using training data captured with setups such as light and camera stages. Such datasets are expensive to acquire, not readily available and do not capture all the rich variations of in-the-wild portrait images. In addition, most supervised approaches only focus on relighting, and do not allow camera viewpoint editing. Thus, they only capture and control a subset of the reflectance field. Recently, portrait editing has been demonstrated by operating in the generative model space of StyleGAN. While such approaches do not require direct supervision, there is a significant loss of quality when compared to the supervised approaches. In this paper, we present a method which learns from limited supervised training data. The training images only include people in a fixed neutral expression with eyes closed, without much hair or background variations. Each person is captured under 150 one-light-at-a-time conditions and under 8 camera poses. Instead of training directly in the image space, we design a supervised problem which learns transformations in the latent space of StyleGAN. This combines the best of supervised learning and generative adversarial modeling. We show that the StyleGAN prior allows for generalisation to different expressions, hairstyles and backgrounds. This produces high-quality photorealistic results for in-the-wild images and significantly outperforms existing methods. Our approach can edit the illumination and pose simultaneously, and runs at interactive rates.

Human pose estimation in realistic world conditions raises multiple challenges such as foreground extraction, background update and occlusion by scene objects. Most of existing approaches were demonstrated in controlled environments. In this paper, we propose a framework to improve the performance of existing tracking methods to cope with these problems. To this end, a robust and scalable framework is provided composed of three main stages. In the first one, a probabilistic occupancy grid updated with a Hidden Markov Model used to maintain an up-to-date background and to extract moving persons. The second stage uses component labelling to identify and track persons in the scene. The last stage uses an hierarchical particle filter to estimate the body pose for each moving person. Occlusions are handled by querying the occupancy grid to identify hidden body parts so that they can be discarded from the pose estimation process. We provide a parallel implementation that runs on CPU and GPU at 4 frames per second. We also validate the approach on our own dataset that consists of synchronized motion capture with a single RGB-D camera data of a person performing actions in challenging situations with severe occlusions generated by scene objects. We make this dataset available online.

This work is related to the development of a marker less system allowing the tracking of elderly people at home. Microsoft Kinect is a low cost 3D camera adapted to the tracking of human movements. We propose a method for making the fusion of the information provided by several Kinects. The observed space is tesselated into cells forming a 3D occupancy grid. We calculate a probability of occupation for each cell of the grid. From this probability we distinguish whether the cells are occupied or not by a static object (wall) or a mobile object (chair, human being). This categorization is realized in real-time using a simple three states HMM. The proposed method for discriminating between mobile and static objects in a room is the main contribution of this paper. The use of HMMs allows to deal with an aliasing problem since mobile objects result in the same observation as static objects. The approach is evaluated in simulation and in a real environment showing an efficient real-time discrimination between cells occupied by mobile objects and cells occupied by static objects.

We present a markerless human motion capture system that estimates the 3D positions of the body joints over time. The system uses a dynamic bayesian network and a factored particle filtering algorithm. In this paper we evaluate the impact of using different observation functions for the bayesian state estimation: chamfer distance, a pixel intersection and finally a pseudo-observation of the subject direction calculated from the previous output of the system. We also compare two methods for the factored generation of the particles. The first one uses a deterministic interval exploration strategy whereas the second one is based on an adaptive diffusion. The capacity of the system to recover after occlusion by obstacles was tested on simulated movements in a virtual scene.