Firstname Lastname

Jian Wang

Max-Planck-Institut für Informatik
Department D6: Visual Computing and Artificial Intelligence
 office: Campus E1 4, Room 117
Saarland Informatics Campus
66123 Saarbrücken
 email: jianwang AT
 phone: +49 681 9325 4044
 fax: +49 681 9325 4099

Research Interests

  • Ego-centric Motion Capture
  • Computer Vision


UnrealEgo: A New Dataset for Robust Egocentric 3D Human Motion Capture.

Hiroyasu Akada  Jian Wang  Soshi Shimada  Masaki Takahashi  Christian Theobalt Christian Theobalt

ECCV, 2022

We present UnrealEgo, i.e. a new large-scale naturalistic dataset for egocentric 3D human pose estimation. UnrealEgo is based on an advanced concept of eyeglasses equipped with two fisheye cameras that can be used in unconstrained environments. We design their virtual prototype and attach them to 3D human models for stereo view capture. We next generate a large corpus of human motions. As a consequence, UnrealEgo is the first dataset to provide in-the-wild stereo images with the largest variety of motions among existing egocentric datasets. Furthermore, we propose a new benchmark method with a simple but effective idea of devising a 2D keypoint estimation module for stereo inputs to improve 3D human pose estimation.

[Project Page]

Estimating Egocentric 3D Human Pose in the Wild with External Weak Supervision.

Jian Wang  Lingjie Liu  Weipeng Xu  Kripasindhu Sarkar  Diogo Luvizon  Christian Theobalt

CVPR, 2022

We present a new egocentric pose estimation method, which can be trained on the new dataset with weak external supervision. Specifically, we first generate pseudo labels for the EgoPW dataset with a spatio-temporal optimization method by incorporating the external-view supervision. The pseudo labels are then used to train an egocentric pose estimation network. To facilitate the network training, we propose a novel learning strategy to supervise the egocentric features with the high-quality features extracted by a pretrained external-view pose estimation model.

[Project Page]

Estimating Egocentric 3D Human Pose in Global Space.

Jian Wang   Lingjie Liu   Weipeng Xu   Kripasindhu Sarkar   Christian Theobalt  

ICCV(Oral), 2021

We present a new method for egocentric global 3D body pose estimation using a single head-mounted fisheye camera. To achieve accurate and temporally stable global poses, a spatio-temporal optimization is performed over a sequence of frames by minimizing heatmap reprojection errors and enforcing local and global body motion priors learned from a mocap dataset. Experimental results show that our approach outperforms state-of-the-art methods both quantitatively and qualitatively.

[PDF] [Supplementary Materials] [Video] [Project Page]

Re-Identification Supervised Texture Generation.

Jian Wang   Yunshan Zhong   Yachun Li   Chi Zhang   Yichen Wei  

CVPR, 2019

In this paper, we propose an end-to-end learning strategy to generate textures of human bodies under the supervision of person re-identification. We render the synthetic images with textures extracted from the inputs and maximize the similarity between the rendered and input images by using the re-identification network as the perceptual metrics. Experiment results on pedestrian images show that our model can generate the texture from a single image and demonstrate that our textures are of higher quality than those generated by other available methods. Furthermore, we extend the application scope to other categories and explore the possible utilization of our generated textures.

[PDF] [Code]

NIL: Learning Nonlinear Interpolants.

Mingshuai Chen   Jian Wang   Jie An   Deepak Kapur   Naijun Zhan  

CADE, 2019

We leverage classification techniques with space transformations and kernel tricks as established in the realm of machine learning, and present a counterexample-guided method named NIL for synthesizing polynomial interpolants, thereby yielding a unified framework tackling the interpolation problem for the general quantifier-free theory of nonlinear arithmetic, possibly involving transcendental functions. We prove the soundness of NIL and propose sufficient conditions under which NIL is guaranteed to converge, i.e., the derived sequence of candidate interpolants converges to an actual interpolant, and is complete, namely the algorithm terminates by producing an interpolant if there exists one.

[PDF] [Code]

From Model to Implementation: A Network-Algorithm Programming Language.

Jian Wang   Jie An   Mingshuai Chen   Naijun Zhan   Lulin Wang  
Miaomiao Zhang   Ting Gan  

SCIENCE CHINA Information Sciences, 2019

We define a novel network algorithm programming language (NAPL) that enhances the SDN framework with a rapid programming flow from topology-based network models to C++ implementations, thus bridging the gap between the limited capability of existing SDN APIs and the reality of practical network management. In contrast to several state-of-the-art languages, NAPL provides a range of critical high-level network programming features: (1) topology-based network modeling and visualization; (2) fast abstraction and expansion of network devices and constraints; (3) a declarative paradigm for the fast design of forwarding policies; (4) a built-in library for complex algorithm implementation; (5) full compatibility with C++ programming; and (6) user-friendly debugging support when compiling NAPL into highly readable C++ codes.


Academic Service

  • Conference refereeing:
    CVPR 2022, ECCV 2022, TASE 2019, RTCSA 2019


  • Summer semester 2022, 2021:
    Supervisor for Computer Vision and Machine Learning for Computer Graphics Seminar, Saarland University and MPI for Informatics