3D human reconstruction has been explored for several decades in the field of computer vision and computer graphics. Accurate methods based on stereo or fusion have been proposed using various types of sensors[12, 44, 33, 35, 40, 49, 50], and several applications have become popular in sports, medicine and entertainment (e.g., movies, games, AR/VR experiences). However, these setups require tightly controlled environments. To date, full 3D human reconstruction with detailed geometry and appearance from in-the-wild pictures is still challenging (i.e., taken in natural conditions as opposed to laboratory environments). Moreover, the lack of automatic rigging prevents animation-based applications.
Recent computer vision models have enabled the recovery of 2D and 3D human pose and shape estimation from a single image. However, they usually rely on representations that have limitations: (1) skeletons  are kinematic structures that are accurate to represent 3D poses, but do not carry body shape information. (2) surface meshes [19, 37, 51] can represent body shape geometry, but have topology constraints; (3) voxels  are topology-free, but memory costly with limited resolution, and need to be rigged for animation. In this paper, we propose the ARCH (Animatable Reconstruction of Clothed Humans) framework that possesses all benefits of current representations. In particular, we introduce a learned model that has human body structure knowledge (i.e., body part semantics), and is trained with humans in arbitrary poses.
First, 3D body pose and shape estimation can be inferred from a single image of a human in arbitrary pose by a prediction model. This initialization step is used for normalized-pose reconstruction of clothed human shape within a canonical space. This allows us to define a Semantic Space (SemS) and a Semantic Deformation Field (SemDF) by densely sampling 3D points around the clothed body surface and assigning skinning weights. We then learn an implicit function representation of the 3D occupancy in the canonical space based on SemS and SemDF, which enables the reconstruction of high-frequency details of the surface (including clothing wrinkles, hair style, etc.) superior to the state of the art [34, 42, 45]. The surface representing a clothed human in a neutral pose is implicitly rigged in order to be used as an animatable avatar. Moreover, a differentiable renderer is used to refine normal and color information for each 3D point in space by Granular Render-and-Compare. Here, we regard them as a sphere and develop a new blending formulation based on the estimated occupancy. See Fig. 2 for an overview of the framework.
In our experiments, we evaluate ARCH on the task of 3D human reconstruction from a single image. Both quantitative and qualitative experimental results show ARCH outperforms state-of-the-art body reconstruction methods on public 3D scan benchmarks and in-the-wild 2D images. We also show that our reconstructed clothed humans can be animated by motion capture data, demonstrating the potential applications for human digitization for animation.
Contributions. The main contributions are threefold: 1) we introduce the Semantic Space (SemS) and Semantic Deformation Field (SemDF) to handle implicit function representation of clothed humans in arbitrary poses, 2) we propose opacity-aware differentiable rendering to refine our human representation via Granular Render-and-Compare, and 3) we demonstrate how reconstructed avatars can directly be rigged and skinned for animation. In addition, we learn per-pixel normals to obtain high-quality surface details, and surface albedo for relighting applications.
2 Related Work
3D clothed human reconstruction focuses on the task of reconstructing 3D humans with clothes. There are multiple attempts to solve this task with video inputs [2, 3, 39, 1, 52], RGB-D data [53, 55] and in multi-view settings [5, 13, 14, 46, 47, 15, 48, 6]. Though richer inputs clearly provide more information than single images, the developed pipelines yield more limitations on the hardware and additional time costs in deployment. Recently, some progress [7, 16, 19, 21, 22, 24, 43, 51, 56] has been made in estimating parametric human bodies from a single RGB image, yet boundaries are under-explored to what extent 3D clothing details can be reconstructed from such inputs. In recent work [23, 25, 4], the authors learn to generate surface geometry details and appearance using 2D UV maps. While details can be learned, the methods cannot reconstruct loose clothing (e.g., dress) and recover complex shapes such as hair or fine structures (e.g., shoe heels). Due to different types of clothing topology, volumetric reconstruction has great benefits in this scenario. For example, BodyNet  takes a person image as input and learns to reconstruct voxels of the person with additional supervision through body priors (e.g., 2D pose, 3D pose, part mask); while PIFu  assumes no body prior and learns an implicit surface function based on aligned image features, leading more clothes details and less robustness against pose variations.
In this paper, we incorporate body prior knowledge to transform people in arbitrary poses to the canonical space, and then learn to reconstruct an implicit representation.
makes the rendering operation differentiable and uses it to optimize parameters of the scene representation. Existing approaches can be roughly divided into two categories: mesh rasterization based rendering[9, 20, 27, 30, 26] and volume based rendering [17, 28]. For example, OpenDR  and Neural Mesh Renderer  manually define approximated gradients of the rendering operation to move the faces. SoftRasterizer  and DIB-R , in contrast, redefine the rasterization as a continuous and differentiable function, allowing gradients to be computed automatically. For volume-based differentiable rendering,  represents each 3D point as a multivariate Gaussian and performs occlusion reasoning with grid discretization and ray tracing. Such methods require an explicit volume to perform occlusion reasoning.  develops differentiable rendering for implicit surface representations with a focus on reconstructing rigid objects.
In contrast, we use a continuous rendering function as in , but revisit it to handle opacity, and we use geometric primitives at points of interest and optimize their properties.
3 Proposed Framework
ARCH contains three components, after 3D body estimation by  (see Fig. 2): pose-normalization using Semantic Space (SemS) and Semantic Deformation Field (SemDF), implicit surface reconstruction, and refinement using a differentiable renderer by Granular Render-and-Compare (see Sec. 3.4).
3.1 Semantic Space and Deformation Field
Our goal is to transform an arbitrary (deformable) object into a canonical space where the object is in a predefined rest pose. To do so, we introduce two concepts: the Semantic Space (SemS) and the Semantic Deformation Field (SemDF). SemS is a space consisting of 3D points where each point is associated to semantic information
enabling the transformation operation. SemDF is a vector field represented by a vector-valued functionthat accomplishes the transformation,
In computer vision and graphics, 3D human models have been widely represented by a kinematic structure mimicking the anatomy that serves to control the pose, and a surface mesh that represents the human shape and geometry. Skinning is the transformation that deforms the surface given the pose. It is parameterized by skinning weights that individually influence body part transformations . In ARCH, we define SemS in a similar form, with skinning weights.
Assuming a skinned body template model in a normalized A-pose (i.e., the rest pose), its associated skeleton in the canonical space, and skinning weights , SemS is then
where each point is associated to a collection of skinning weights defined with respect to body parts (e.g., skeleton bones). In this paper, we approximate by retrieving the closest point on the template surface to and assigning the corresponding skinning weights from . In practice, we set a distance threshold to cut off points that are too far away from .
In ARCH, SemDF actually performs an inverse-skinning transformation, putting a human in arbitrary pose to its normalized-pose in the canonical space. This extends standard skinning (e.g., Linear Blend Skinning or LBS ) applied to structured objects to arbitrary 3D space and enables transforming an entire space in arbitrary poses to the canonical space, as every point can be expressed as a linear combination of points with skinning weights .
Following LBS, the canonical space of human body is tied to a skeletal rig. The state of the rig is described by relative rotations of all skeleton joints . Every rotation is relative to the orientation of the parent element in a kinematic tree. For a skeleton with body parts, . Given a body template model in rest pose with vertices, the LBS function takes as input the vertices , the joints , a target pose , and deforms every to the posed position with skinning weights , namely,
where is the rest-pose corrected affine transformation to apply to body part .
3.2 Implicit Surface Reconstruction
We use the occupancy map to implicitly represent the 3D clothed human, i.e.,
where denotes the occupancy for a point . To obtain a surface, we can simply threshold the occupancy map to obtain the isosurface .
In this paper, we incorporate a human body prior by always reconstructing a neutral-posed shape in the canonical space. Similar to 
, we develop a deep neural network that takes a canonical space point, its correspondent 2D position , and the 2D image as inputs and estimates occupancy , normal , color for ; that is,
where , and denote the occupancy, normal and color sub-network weights, is the spatial feature extracted based on SemS. We use the estimated canonical body landmarks from 
and compute the Radial Basis Function (RBF) distance betweenand the -th landmark , that is
where is the Euclidean distance. We also evaluate the effects of different types of spatial features in Sec. 4.3. and the feature maps extracted from occupancy and normal sub-networks, respectively (see also Fig. 2). The three sub-networks are defined as follows:
The Occupancy sub-network uses a Stacked Hourglass (SHG) 
as the image feature encoder and a Multi-Layer Perceptron (MLP) as the regressor. Given ainput image , the SHG produces a feature map with the same grid size. For each 3D point , we consider the feature located at the corresponding projected pixel as its visual feature descriptor
. For points that do not align onto the grid, we apply bi-linear interpolation on the feature map to obtain the feature at that pixel-aligned location. The MLP takes the spatial feature of the 3D pointand the pixel-aligned image features as inputs and estimates the occupancy
by classifying whether this point lies inside the clothed body or not.
The Normal sub-network uses a U-net  as the image feature encoder and a MLP which takes the spatial feature, and feature descriptors and from its own backbone and from the occupancy sub-network as inputs and estimates the normal vector .
The Color sub-network also uses a U-net  as the image feature encoder and a MLP which takes the spatial feature, and feature descriptors , and from its own backbone, as well as the normal and occupancy sub-networks as inputs and estimates the color in RGB space.
For each sub-network, the MLP takes the pixel-aligned image features and the spatial features (as described in Sec. 3.1
), where the numbers of hidden neurons are. Similar to , each layer of MLP has skip connections from the input features. For the occupancy sub-network, the MLP estimates one-dimension occupancy using Sigmoid activation. For the normal sub-network, the MLP estimates three-dimension normal using L2 normalization. For the color sub-network, the MLP estimates three-dimension color using range clamping.
During training, we optimize the parameters of all three sub-models, i.e
., the occupancy, normal and color models. We define the training in three separate loops to train each part with the appropriate losses and avoid computational bottlenecks. The total loss function is defined as
where is the 3D loss for occupancy network, and are the 3D and 2D losses for normal network, and and are the 3D and 2D losses for color network. For every training iteration, we perform the following three optimizations.
Occupancy. We use the available ground truth to train the occupancy prediction model in a direct and supervised way. First, we sample. This turned out to cover the various body shapes and clothing well in our experiments, but can be selected according to the data distribution at hand. These points are then processed by the occupancy model, providing us with an estimated occupancy value for every sampled point. We use a sigmoid function on these values to normalize the network output to the interval , where we select as the position of the isosurface. is the position where the derivative of the sigmoid function is the highest and we expect to optimize the surface prediction best. The loss is defined as the Huber loss comparing the occupancy prediction and ground truth. Similar to , we found a less aggressive loss function than the squared error better suited for the optimization, but found the quadratic behavior of the Huber loss around zero to be beneficial.
Normals and colors for surface points. Colors and normals can be optimized directly from the ground truth mesh for points that lie on its surface. To use this strong supervision signal we introduce a dedicated training stage. In this stage, we sample points only from the mesh surface and push them through the color and normal models. In our setup, we use point samples per model per training step. The loss terms and are defined as the loss comparing the predicted normals and colors with the ground truth across all surface points. The occupancy predictions are kept unchanged.
Normals and colors for points not on the surface. For points not on the mesh surface, it is not clear how the ground truth information can be used in the best way to improve the prediction without an additional mapping. In a third step for the training, we sample another set of points, and push them through the occupancy, color and normal models and use a differentiable renderer on the prediction. We render the image using the occupancy information as opacity, and by using the color channels to represent colors or normals and use the gradients to update the predicted values. and are defined as the per-pixel L1 loss between the rendered image and the ground truth. For details on this step, see Fig. 3 and the following Sec. 3.4.
3.4 Granular Render-and-Compare
The prediction from the model is an implicit function representation. By sampling points in a predefined volume and optimizing , and , we can optimize the occupancy, normal and color at these points directly given 3D ground truth. However, it is not clear what the gradients should be for points that are located not directly on the surface of the ground truth mesh. To address this problem, we propose to use a differentiable renderer.
We first create an explicit geometric representation of the scene at hand. For every sample point to optimize, we place a geometric primitive with a spatial extent at its position. To be independent of the viewpoint, we choose this to a sphere with radius for every sampled point (for an overview of the differentiable rendering loss computation, see Fig. 3). During training, every scene to render contains spheres.
We then define a differentiable rendering function  to project the spheres onto the image plane so that we can perform pixel-level comparisons with the projected ground truth. We use a linear combination with a weight to associate the color contribution from point to the pixel . Having the color and normal for point , the color and normal for pixel are calculated as the weighed linear combination of point values and .
We define considering two factors: the depth of the sphere for point at pixel , , and the proximity of the projected surface of the sphere for point to pixel , . To make occlusion possible, the depth needs to have a strong effect on the resulting weight. Hence,  defines the weight as
with being a small numerical constant. With this definition, the proximity has linear influence on the resulting weight while the depth has exponential influence. The impact ratio is controlled by the scaling factor , which we fix to in our experiments.
In contrast to  we also need to use an opacity per sphere for rendering. We tie this opacity value directly to the predicted occupancy value through linear scaling and shifting. To stay with the formulation of the render function, we integrate into the weight formulation in Eqn. 7.
If the opacity is used as a linear factor in this equation, the softmax function will still render spheres with very low opacity over other spheres with a lower depth value. The problem is the exponential function that is applied to the scaled depth values. On the other hand, if an opacity factor is only incorporated into the exponential function, spheres will remain visible in front of the background (their weight factor is still larger than the background factor ). We found a solution by using the opacity value as both, linear scaling factor as well as exponential depth scaling factor. This solution turned out to be numerically stable and well-usable for optimization with all desired properties. This changes the weight function to the following:
Using this formulation, we optimize the color channel values and normal values per point. A per-pixel L1 loss is computed between the rendering and a rendering of the ground truth data and back-propagated through the model. For our experiments with and the depth of the volume, we map the occupancy values that define the isosurface at the value to the threshold where shifts to transparency. We experimentally determined this value to be roughly .
For inference, we take as input a single RGB image representing a human in an arbitrary pose, and run the forward model as described in Sec. 3.2 and Fig. 2. The network outputs a densely sampled occupancy field over the canonical space from which we use the Marching Cube algorithm  to extract the isosurface at threshold . The isosurface represents the reconstructed clothed human in the canonical pose. Colors and normals for the whole surface are also inferred by the forward pass and are pixel-aligned to the input image (see Sec. 3.2). The human model can then be transformed to its original pose by LBS using SemDF and per-point corresponding skinning weights as defined in Sec. 3.1.
Furthermore, since the implicit function representation is equipped with skinning weights and skeleton rig, it can naturally be warped to arbitrary poses. The proposed end-to-end framework can then be used to create a detailed 3D avatar that can be animated with unseen sequences from a single unconstrained photo (see Fig. 5).
We present details on ARCH implementation and datasets for training, with results and comparisons to the state of the art.
4.1 Implementation Details
ARCH is implemented in PyTorch. We train the neural network model using the RMSprop optimizer with a learning rate starting from 1e-3. The learning rate is updated using an exponential schedule every 3 epochs by multiplying with the factor. We are using 3D scans to train the model and use views per epoch, resulting in images for the training per epoch. Training the model on an NVIDIA DGX-1 system with one Tesla V100 GPU takes for 9 epochs.
Our training dataset is composed of 375 3D scans from the RenderPeople111http://renderpeople.com dataset, and 207 3D scans from the AXYZ222http://secure.axyz-design.com dataset. The scans are watertight meshes which are mostly free of noise. They represent subjects wearing casual clothes, and potentially holding small objects (e.g., mobile phones, books and purses). Our test dataset contains 64 scans from the RenderPeople dataset, 207 scans from the AXYZ dataset, 26 scans from the BUFF dataset , and 2D images from the DeepFashion  dataset, representing clothed people with a large variety of complex clothing. The subjects in the training dataset are mostly in standing pose, while the subjects in the test dataset are in arbitrary poses (standing, bending, sitting, …). We create renders of the 3D scans using Blender. For each 3D scan, we produce 360 images by rotating a camera around the vertical axis with intervals of 1 degree. For the current experiments, we only considered the weak perspective projection (orthographic camera) but this can be easily adapted. We also used 38 environment maps to render each scan with different natural lighting conditions. The proposed model is trained to predict albedo (given by ground truth scan color). We also observed that increasing the number of images improves the fidelity of predicted colors (as in ).
In order to use a 3D scan for model training, we fit a rigged 3D body template to the scan mesh to estimate the 3D body pose (see Fig. 4). The estimated parametric 3D body can directly serve as ground truth input data during the model training step (see Sec. 3.3). This also allows us to obtain SemS and SemDF for the scan. However, since each 3D scan has its own topology, artifacts due to topology changes will occur when pose-normalization is naively applied to models containing self-contact (for example arms touching the body). This creates inaccurate deformations. Hence, we first detect regions of self-contact and topology changes and cut the mesh before pose-normalization (see Fig. 4 (c) and (d)). Holes are then filled up using Smooth Signed Distance Surface reconstruction  (see Fig. 4 (c) and (d)). For inference on 2D images from the DeepFashion dataset, we obtain 3D body poses using the pre-trained models from .
4.3 Results and Comparisons
We evaluate the reconstruction accuracy of ARCH with three metrics similar to . We reconstruct the results on the same test set and repose them back to the original poses of the input images and compare the reconstructions with the ground truth surfaces in the original poses. We report the average point-to-surface Euclidean distance (P2S) in centimeters, the Chamfer distance in centimeters, and the L2 normal re-projection error in Tab. 1.
Additionally to comparing with state-of-the-art methods [10, 18, 19, 34, 42, 45], we include scores of an ablative study with the proposed method. In particular, we evaluate three variants and validate the effectiveness of two main components: the Semantic Deformation Field and the Granular Render-and-Compare loss.
ARCH, baseline: a variant of  using our own network specifications, taking an image as input and directly estimating the implicit surface reconstruction.
Semantic Deformation Field (SemDF): we first estimate the human body configuration by  and then reconstruct the canonical shape using the implicit surface reconstruction, and finally repose the canonical shape to the original pose in the input image.
Granular Render-and-Compare (GRaC): based on the previous step, we further refine the reconstructed surface normal and color using differentiable render-and-compare.
ARCH baseline specification already achieves state-of-the-art performance in normal estimation, but has inferior performance w.r.t. P2S and Chamfer error compared to PIFu . We use a different training dataset compared to PIFu that apparently does not represent the test set as well. Also, PIFu normalizes every scan at training and prediction time to have its geometric center at the coordinate origin, whereas we use origin placed scans with slight displacements. Lastly, PIFu performs a size normalization of the body using the initial 3D body configuration estimate. The image is rescaled so that the height of the person matches the canonical size. This makes person height estimation for PIFu impossible, whereas we properly reconstruct it—at the cost of a more difficult task to solve. The benefit of this operation is not reflected in the scores because the metrics are calculated in the original image space.
When adding SemDF, we see a substantial gain in performance compared to our own baseline, but also to the so far best-performing PIFu metrics. We outperform PIFu on average with an improvement of over 50% on the RenderPeople dataset and an average improvement of over 60% on the BUFF dataset. When adding the Granular Render-and-Compare loss, these number improve again slightly, especially on the normal estimation. Additionally, the results gain a lot of visual fidelity and we manage to remove a lot of visual artifacts.
Fig. 7 shows the level of detail of geometry, normal and color predictions our model can achieve. Note that, for example, the zipper is not reproduced in the predicted normal map. This is an indicator that the model does not simply reproduce differences in shading directly in the normal map, but is able to learn about geometric and shading properties of human appearance. In Fig. 6, we show qualitative results on challenging poses from the BUFF dataset. In Fig. 9, we provide a comparison of results of our method with a variety of state of the art models [45, 19, 42].
|Spatial Feature Types||Normal||P2S||Chamfer|
Ablative Studies. We evaluate the effectiveness of different types of spatial features in Tab. 2 and Fig. 8. We evaluate three different features: XYZ uses the absolute position of the sampled point, L2 uses the Euclidean distance from the sampled point to each body landmark, and RBF denotes our proposed method in Sec. 3.1. It can be observed that RBF feature works best for this use case both qualitatively and quantitatively. RBF features strongly emphasize features that are close in distance to the currently analyzed point and puts less emphasis on points further away, facilitating optimization and preserving details.
Animating Reconstructed Avatars. With the predicted occupancy field we can reconstruct a mesh that is already rigged and can directly be animated. We show the animation of an avatar we reconstructed from the AXYZ dataset in Fig. 5, driven by an off-the-shelf retargetted Mixamo animation . By working in the canonical space, the avatar is automatically rigged and can be directly animated. Given only a single view image, the avatar is reconstructed in 3D and looks plausible from all sides.
As shown in Fig 10, rare poses not sufficiently covered in the training dataset (e.g., kneeing) return inaccurate body prior, and are then challenging to reconstruct. Also, details (i.e., normals) in occluded areas could be improved with specific treatment of occlusion-aware estimation.
In this paper, we propose ARCH, an end-to-end framework to reconstruct clothed humans from unconstrained photos. By introducing the Semantic Space and Semantic Deformation Field, we are able to handle reconstruction from arbitrary pose. We also propose a Granular Render-and-Compare loss for our implicit function representation to further constrain visual similarity under randomized camera views. ARCH shows higher fidelity in clothing details including pixel-aligned colors and normals with a wider range of human body configurations. The resulting models are animation-ready and can be driven by arbitrary motion sequences. We will explore handling heavy occlusion cases with in-the-wild images in the future.
Acknowledgements. We would like to thank Junbang Liang and Yinghao Huang (Interns at FRL) for their work on dataset creation.
Learning to reconstruct people in clothing from a single RGB camera.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Cited by: §2.
-  (2018) Detailed human avatars from monocular video. In International Conference on 3D Vision, Cited by: §2.
-  (2018) Video based reconstruction of 3d people models. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2019-10) Tex2Shape: detailed full human body geometry from a single image. In IEEE International Conference on Computer Vision (ICCV), Cited by: §2.
-  (2007) Detailed human shape and pose from images. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2019-10) Multi-garment net: learning to dress 3d people from images. In IEEE International Conference on Computer Vision (ICCV), Cited by: §2.
-  (2016) Keep it SMPL: automatic estimation of 3d human pose and shape from a single image. In European Conference on Computer Vision, Cited by: §2.
-  (2011) SSD: smooth signed distance surface reconstruction. Comput. Graph. Forum 30 (7), pp. 1993–2002. Cited by: §4.2.
-  (2019) Learning to predict 3d objects with an interpolation-based differentiable renderer. In Annual Conference on Neural Information Processing Systems, Cited by: §2.
-  (2019) Learning implicit fields for generative shape modeling. IEEE Conference on Computer Vision and Pattern Recognition. Cited by: §4.3, Table 1.
Learning pose grammar to encode human body configuration for 3d pose estimation.
AAAI Conference on Artificial Intelligence, Cited by: §1.
-  (2007) Accurate, dense, and robust multi-view stereopsis. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §1.
-  (2010) Accurate, dense, and robust multiview stereopsis. IEEE Transactions on Pattern Analysis and Machine Intelligence 32 (8), pp. 1362––1376. Cited by: §2.
-  (2009) Motion capture using joint skeleton tracking and surface estimation. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2000) Image-based visual hulls. In ACM SIGGRAPH, Cited by: §2.
-  (2019) HoloPose: holistic 3d human reconstruction in-the-wild. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2018) Unsupervised learning of shape and pose with differentiable point clouds. In Annual Conference on Neural Information Processing Systems, Cited by: §2.
-  (2018) 3D human body reconstruction from a single image via volumetric regression. European Conference of Computer Vision Workshops. Cited by: §4.3, Table 1.
-  (2018) End-to-end recovery of human shape and pose. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §1, §2, Figure 9, §4.3, §4.3.
-  (2018) Neural 3d mesh renderer. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2019) Learning to reconstruct 3d human pose and shape via model-fitting in the loop. In IEEE International Conference on Computer Vision, Cited by: §2.
-  (2019) Convolutional mesh regression for single-image human shape reconstruction. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2018) DeepWrinkles: accurate and realistic clothing modeling. In European Conference on Computer Vision, Cited by: §2.
-  (2017) Unite the people: closing the loop between 3d and 2d human representations. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2019) 360-degree textures of people in clothing from a single image. In International Conference on 3D Vision, Cited by: §2.
-  (2018) Differentiable monte carlo ray tracing through edge sampling. ACM Transactions on Graphics 37 (6), pp. 222:1–222:11. Cited by: §2.
-  (2019) Soft rasterizer: a differentiable renderer for image-based 3d reasoning. IEEE International Conference on Computer Vision. Cited by: §2, §2, §3.4, §3.4, §3.4.
-  (2019) Learning to infer implicit surfaces without 3d supervision. Annual Conference on Neural Information Processing Systems. Cited by: §2.
-  (2016) Deepfashion: powering robust clothes recognition and retrieval with rich annotations. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §4.2.
-  (2014) OpenDR: an approximate differentiable renderer. In European Conference on Computer Vision, Cited by: §2.
-  (2015) SMPL: a skinned multi-person linear model. ACM Transactions on Graphics 34 (6), pp. 248. Cited by: §3.1, §3.1.
-  (1987) Differentiable monte carlo ray tracing through edge sampling. ACM SIGGRAPH Computer Graphics 21 (4), pp. 163–169. Cited by: §3.5.
-  (2012) 3D video and its applications. Springer. Cited by: §1.
-  (2019) SiCloPe: silhouette-based clothed people. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §1, §4.3, Table 1.
-  (2015) DynamicFusion: reconstruction and tracking of non-rigid scenes in real-time. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §1.
-  (2016) Stacked hourglass networks for human pose estimation. In European Conference on Computer Vision, Cited by: §3.2.
Neural body fitting: unifying deep learning and model-based human pose and shape estimation. In International Conference on 3D Vision, Cited by: §1.
-  (2019) DeepSDF: learning continuous signed distance functions for shape representation. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Cited by: §3.3.
-  (2017) ClothCap: seamless 4d clothing capture and retargeting. ACM Transactions on Graphics 36 (4), pp. 73:1––73:15. Cited by: §2.
-  (2018) Scene-centric joint parsing of cross-view videos. In AAAI Conference on Artificial Intelligence, Cited by: §1.
-  (2015) U-net: convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer-Assisted Intervention, Cited by: §3.2, §3.2.
-  (2019) PIFu: pixel-aligned implicit function for high-resolution clothed human digitization. In IEEE International Conference on Computer Vision, Cited by: §1, §2, §3.2, §3.2, Figure 6, Figure 9, §4.2, §4.3, §4.3, §4.3, §4.3, §4.3, Table 1.
Self-supervised learning of motion capture. In Annual Conference on Neural Information Processing Systems, Cited by: §2.
-  (2009) Complete multi-view reconstruction of dynamic scenes from probabilistic fusion of narrow and wide baseline stereo. In IEEE 12th International Conference on Computer Vision ICCV, Cited by: §1.
-  (2018) BodyNet: volumetric inference of 3D human body shapes. In European Conference on Computer Vision, Cited by: §1, §1, §2, Figure 9, §4.3, §4.3, Table 1.
-  (2008) Articulated mesh animation from multi-view silhouettes. ACM Transactions on Graphics 27 (3), pp. 97:1––97:9. Cited by: §2.
-  (2009) Dynamic shape capture using multi-view photometric stereo. In ACM SIGGRAPH, Cited by: §2.
-  (2012) Full body performance capture under uncontrolled and varying illumination: a shading-based approach. In European Conference on Computer Vision, Cited by: §2.
-  (2016) Multi-view people tracking via hierarchical trajectory composition. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §1.
-  (2017) Multi-view people tracking via hierarchical trajectory composition. In AAAI Conference on Artificial Intelligence, Cited by: §1.
-  (2019) DenseRaC: joint 3D pose and shape estimation by dense render-and-compare. In IEEE International Conference on Computer Vision, Cited by: §1, §1, §2, §3.2, §3, §4.2, §4.3, §4.3.
-  (2016) Estimation of human body shape in motion with wide clothing. In European Conference on Computer Vision, Cited by: §2.
-  (2018) DoubleFusion: real-time capture of human performances with inner body shapes from a single depth sensor. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §2.
-  (2017) Detailed, accurate, human shape estimation from clothed 3D scan sequences. In IEEE Conference on Computer Vision and Pattern Recognition, Cited by: §4.2.
-  (2018) HybridFusion: real-time performance capture using a single depth sensor and sparse imus. In European Conference on Computer Vision, Cited by: §2.
-  (2019) DeepHuman: 3d human reconstruction from a single image. In IEEE International Conference on Computer Vision, Cited by: §2.