Ziyan Wang , Undergraduate

Tsinghua University, Beijing 100084, China

Birthday 1994-11-15

Tel: +86 188 1176 3608

Email: zy-wang13@mails.tsinghua.edu.cn

[Curriculum Vitae]

I am currently a senior undergraduate student in Department of Automation, School of Information Science and Technology, Tsinghua University.

From Winter 2014 to Fall 2015, I worked at Institute of Control Theory and Technology, Broadband Network & Digital Media Lab(BBNC), supervised by Prof. Jingli Suo. Since Winter 2015, I have been working in Institute of Information Processing, Intelligent Vision Group(IVG), advised by Prof. Jiwen Lu. In summer 2016, I visited University of Michigan, Ann Arbor and worked at Vision & Intelligent Lab, supervised by Prof. Jia Deng.

My research interest lie in computer vision, machine learning and deep learning.

Aug. 2013 - Jul. 2017 (Expected), Department of Automation, Tsinghua University

GPA:92/100         Ranking: 7/145

Jun. 2016 - Sep. 2016 , Computer Science and Engineering, University of Michigan, Ann Arbor

Multi-Modal Deep Learning on RGB-D Object Recognition

In this project, I developed a Multi-Modal Deep Learning algorithm which aims to learn a more discriminitive feature for RGB-D object. Our correlated and individual multi-modal deep learning jointly learns feature representations from raw RGB-D data with a pair of residual networks, in which way the shareable and modal-specific information can be simutaneously and explicitly exploited. We achieved better recognition accuracy on two RGB-D dataset: RGB-D object dataset(92.4%) and 2D3D dataset(93.8%). This work is submitted to IEEE Conference on Computer Vision and Pattern Recogntion 2017.

Multiple Object Tracking using Stacked Hourglass Networks

Multi-Object Tracking(MOT) is an interesting visual task and can be applied to a lot of real world applications. Given the most popular MOT criteria, "tracking by detection", we developed an algorithm that is able to link detection and track by learning tracking strategy. Considering that the targets are often suffered from severe articulation, occlusion, deformation and change in appearance, we adopt an hourglass network to learn the representation for relationship between detection and tracks which share a lot of common between different targets. We evaluate our method on MOT2015 and got a 5-10% promotion on MOT accuracy on validation set comparing to the state-of-the-art method. This work is submitted to IEEE Conference on Computer Vision and Pattern Recogntion 2017.

1. Ziyan Wang, Jiwen Lu, Ruogu Lin, Jianjiang Feng, Jie Zhou, "Correlated and Individual Multi-Modal Deep Learning for RGB-D Object Recognition", submitted to IEEE Conference on Computer Vision and Pattern Recognition, 2017

2. Hei Law, Xuchen You, Ziyan Wang, Jia Deng, "Multiple Object Tracking using Stacked Hourglass Networks", submitted to IEEE Conference on Computer Vision and Pattern Recognition, 2017