In this paper, we propose a novel approach to reconstruct 3D human body shapes based on a sparse set of RGBD frames using a single RGBD camera. We specifically focus on the realistic settings where human subjects move freely during the capture. The main challenge is how to robustly fuse these sparse frames into a canonical 3D model, under pose changes and surface occlusions. This is addressed by our new framework consisting of the following steps. First, based on a generative human template, for every two frames having sufficient overlap, an initial pairwise alignment is performed; It is followed by a global non-rigid registration procedure, in which partial results from RGBD frames are collected into a unified 3D shape, under the guidance of correspondences from the pairwise alignment; Finally, the texture map of the reconstructed human model is optimized to deliver a clear and spatially consistent texture. Empirical evaluations on synthetic and real datasets demonstrate both quantitatively and qualitatively the superior performance of our framework in reconstructing complete 3D human models with high fidelity. It is worth noting that our framework is flexible, with potential applications going beyond shape reconstruction. As an example, we showcase its use in reshaping and reposing to a new avatar.

This paper presents a novel framework to recover detailed avatar from a single image. It is a challenging task due to factors such as variations in human shapes, body poses, texture, and viewpoints. Prior methods typically attempt to recover the human body shape using a parametric-based template that lacks the surface details. As such resulting body shape appears to be without clothing. In this paper, we propose a novel learning-based framework that combines the robustness of the parametric model with the flexibility of free-form 3D deformation. We use the deep neural networks to refine the 3D shape in a Hierarchical Mesh Deformation (HMD) framework, utilizing the constraints from body joints, silhouettes, and per-pixel shading information. Our method can restore detailed human body shapes with complete textures beyond skinned models. Experiments demonstrate that our method has outperformed previous state-of-the-art approaches, achieving better accuracy in terms of both 2D IoU number and 3D metric distance.

This paper deals with the 3D reconstruction problem for dynamic non-rigid objects with a single RGB-D sensor. It is a challenging task as we consider the almost inevitable accumulation error issue in some previous sequential fusion methods and also the possible failure of surface tracking in a long sequence. Therefore, we propose a global non-rigid registration framework and tackle the drifting problem via an explicit loop closure. Our novel scheme starts with a fusion step to get multiple partial scans from the input sequence, followed by a pairwise non-rigid registration and loop detection step to obtain correspondences between neighboring partial pieces and those pieces that form a loop. Then, we perform a global registration procedure to align all those pieces together into a consistent canonical space as guided by those matches that we have established. Finally, our proposed model-update step helps fixing potential misalignments that still exist after the global registration. Both geometric and appearance constraints are enforced during our alignment; therefore, we are able to get the recovered model with accurate geometry as well as high fidelity color maps for the mesh. Experiments on both synthetic and various real datasets have demonstrated the capability of our approach to reconstruct complete and watertight deformable objects.

The goal of human-robot motion retargeting is to let a robot follow the movements performed by a human subject. Typically in previous approaches, the human poses are precomputed from a human pose tracking system, after which the explicit joint mapping strategies are specified to apply the estimated poses to a target robot. However, there is not any generic mapping strategy that we can use to map the human joint to robots with different kinds of configurations. In this paper, we present a novel motion retargeting approach that combines the human pose estimation and the motion retargeting procedure in a unified generative framework without relying on any explicit mapping. First, a 3D parametric human-robot ( HUMROB ) model is proposed which has the specific joint and stability configurations as the target robot while its shape conforms the source human subject. The robot configurations, including its skeleton proportions, joint limitations, and DoFs are enforced in the HUMROB model and get preserved during the tracking procedure. Using a single RGBD camera to monitor human pose, we use the raw RGB and depth sequence as input. The HUMROB model is deformed to fit the input point cloud, from which the joint angle of the model is calculated and applied to the target robots for retargeting. In this way, instead of fitted individually for each joint, we will get the joint angle of the robot fitted globally so that the surface of the deformed model is as consistent as possible to the input point cloud. In the end, no explicit or pre-defined joint mapping strategies are needed. To demonstrate its effectiveness for human-robot motion retargeting, the approach is tested under both simulations and on real robots which have a quite different skeleton configurations and joint degree of freedoms (DoFs) as compared with the source human subjects.