CV

Education

• Research Topic: Joint Reconstruction of Hands and Objects.

• GPA: 3.75/ 4.00.
• Master’s Thesis Project: Stereoscopic 3D Human Pose Estimation via Structurally Constrained Optimization.

• Average Credit Point: 89.09/100. Ranking: 9/107
• Bachelor Thesis Project: Intelligent Interactive Design for Exercise Rehabilitation Based on 3D Human Pose Estimation.

Publications

• Zhuo Chen, Xu Zhao, and Xiaoyue Wan. Structural triangulation: A closed-form solution to constrained 3d human pose estimation. In European Conference on Computer Vision (ECCV), Sept 2022. (Published)
• Zhuo Chen and Xu Zhao. Incorporating Limb Orientation Fields to Improve Stereo-View Human Pose Estimation[J]. IEEE Transactions on Multimedia, 2023. (Preprint)
• Xiaoyue Wan, Xu Zhao, and Zhuo Chen. View Consistency Aware Holistic Triangulation for 3D Human Pose Estimation[J]. Computer vision and image understanding, 2023. （Accepted)

Research and Project Experience

Aid exercise rehabilitation using pose estimation from short-baselined binocular vision.

Research Assistant

• Designed and Implemented the bi-directional conversion between joint positions and BVH data.
• Proposed virtual joints to ensure the uniqueness in the above converting process.
• Proposed Limited Pictorial Structure Model (LPSM) for short-baseline binocular pose estimation, with an ~87× speed up in MATLAB compared to PSM.

Collect a motion capture dataset with novel camera settings, multi-module labels, and rehabilitation-aimed moves.

Research Assistant

• Designed collecting scenarios, including the largest possible common visible range of all cameras inside the capture room and the corresponding camera setting.
• Managed labelling process using Vicon motion capture system, including marker set preparation, motion capture process, and label data post-processing.

Accelerate OpenCV functions using Halide framework.

Software Engineer

• Reimplemented imresize function and achieved ~15× average speed up.
• Reimplemented KNN-search algorithm and revealed current drawbacks of Halide in implementing tree structures.
• Met with Huawei staff regularly reporting progress and advising for promoting Halide framework.

  Teaching Experience

Academic Services

Conference Reviewer:

• NeurIPS 2023, 2024;
• ECCV 2022, 2024;
• CVPR 2022

Related Skills

• Python: Fully master ML coding and DNN & parallelism implementation using PyTorch.
• C++: Skilled in coding and compiling.
• MATLAB: Skilled in ML experiments.
• Vicon capture system: Capable of manipulating Nexus in capturing label data and post-processing.