Wenshuang Song

(宋文爽)
3D Vision Algorithm Engineer
Renmin University of China
3D Computer Vision
Email: wenshuangsong@yahoo.com
Tel: 13651363327
Github: https://github.com/WenshuangSong
CSDN: https://blog.csdn.net/sunshine_010

Short Bio

My research interests include 3D computer vision and computer graphics (e.g., 3D human pose and shape reconstruction, 6DoF pose estimation) .

If you are interested in my topics , please send me an email.

Education

Renmin University of China
Bachelor's Degree, Business Management
From Sept. 2013 to Jul. 2017

Research

• Computer graphics
• 3D Computer Vision

Professional Skill

• Familiar with common productive tools, including: Linux, Git, SSH, Docker etc.;
• Familiar with Python and C；
• Familiar with popular deep learning frameworks, such as PyTorch and Tensorflow；

Intern Experience

• Baidu Times Network Technology (Beijing) Co., Ltd. | 2021.03 -- 2021.08

  Responsible for the algorithm development and performance improvement of the Real-Time Augmented Reality Shoe Try-on System. Research directions include but are not limited to: shoes detection, shoes classification, keypoints detection and segmentation and other algorithms.

• Intel (China) Research Center Co., Ltd. | 2020.07 -- 2021.03

  Responsible for main 3D human body reconstruction algorithm, and conduct algorithm research and development in multiple task areas:3D human body pose estimation, 3D human body reconstruction； Develop and implement state-of-art algorithms based on computer graphics and the principles of 3D reconstruction and rendering.

Competition

• Kaggle Target Detection Competition: Global Wheat Detection, to develop the wheat models in different environments，use YOLOV5 and pseudo label to make the best Public Score: 0.7634, Public Leaderboard: Top 5%；
• Kaggle Image Classification Algorithm: classify objects in 10 natural scenes of CIFAR10, with an accuracy rate of 94%；

Research Results

• Self-supervised learning method of human body shape and pose estimation in monocular video based on skeleton constraints. [Patent] | 2021.09
• Video inference for human body pose and shape estimation based on temporal information. [Patent] | 2021.09
• A trajectory constraint based optimization method for human body pose and shape reconstruction. [Patent] | 2021.09
• Software copyright of Real-time Iris Segmentation System under unrestricted conditions, 2019