- M.S. in Electrical & Computer Engineering, Sep.2017 - Dec. 2018(expected)
- Member of Information Processing Lab where the main research areas are computer vision and image processing
- B.S. in Electrical & Computer Engineering, Sep. 2013 - Jun. 2017
- Member of IVM Lab, Outstanding Graduate of SJTU
- Interests: Computer Vision, Machine Learning, Deep Learning, Autonomous Driving
- language: C/C++, Python, Java, Matlab
- Tools: OpenCV, Linux, Git, Vim, Bash, TensorFlow, Azure, AWS
Machine Learning Intern in Perception Team
Built an online multi-sensor tracking algorithm, which includes Lidar and cameras, that runs in real-time for an autonomous truck
Utilized lane detection results and 3d map to do real-time camera pose estimation and tracking stabilization
Created a regression test pipeline for better analyzing tracking results of different versions of tracking algorithm
Machine Learning & Computer Vision Intern
- Refactored the model training setup by utilizing multiprocessing pool to reduce time at assembling data and cropping images
- Improved classification accuracy by using super-resolution technique to do fine-grained classification (more than 2000 types)
- Trained an end-to-end fine-grained classification model by combining image super-resolution and classification into a single model
Information Processing Lab (IPL), UW
Graduate Research Assistant, supervised by Prof. Jenq-Neng Hwang
- Realized a fully unsupervised online learning framework to achieve multi-camera tracking of people
- Designed multi-camera tracking of vehicle with a fusion of adaptive appearance, semantic features and comparison of license plates
- Participated in NVIDIA AI City challenge 2018, which held as a workshop at CVPR 2018, and achieved a superiority performance in both Track 1: Traffic Flow Analysis and Track 3: Multi-camera Vehicle Detection & Reidentification, among over 20 teams
Image, Video, and Multimedia Communication Lab (IVM), SJTU
Undergraduate Research Assistant, supervised by Prof. Weiyao Lin.
- Collected two challenging group re-identification datasets by tracking people in a crowd scene and implementing socially constraint structure learning to detect groups.
- Developed a multi-grain group re-identification process which derives features for multi-grain objects and iteratively evaluates their importance to handle interferences from group dynamics.
- Proposed a multi-order matching process by a personalized random walk scheme through a multi-order association graph, which integrated multi-grain information to obtain more reliable group matching results.
Project Leader of an Intelligent System
- Achieved a real-time object detection system by training convolutional neural network and iteratively advancing model performance, which can classify normal people and fallen people.
- Utilized graph matching algorithm based on confident tracklets to develop multiple object tracking algorithm.
- Implemented a real-time tracking system on surveillance video stream by combining detection and tracking algorithm, which realized pedestrian counting as well as fall warning.
Research Center of Intelligent Internet of Things (IIOT), SJTU
Project Team Leader, supervised by Prof. Xinbing Wang.
- Conducted the construction and maintenance of an academic search system: Acemap.
- Analyzed a large scale academic dataset, Microsoft Academic Graph (MAG), which included data crawling, cleaning and processing.
- Applied for a patent: Construction and Visualization of Heterogeneous Topic Web Based on Text Network, which can recognize the relationship between word topics and document topics.
- Designed a novel “interactive map” approach to visualize large-scale and high-dimensional academic literatures and display the underlying relationship among them, which is available on the Acemap website.
Single-camera and inter-camera vehicle tracking and 3D speed estimation based on fusion of visual and semantic features
Zheng Tang, Gaoang Wang, Hao Xiao, Aotian Zheng and Jenq-Neng Hwang, "Single-camera and inter-camera vehicle tracking and 3D speed estimation based on fusion of visual and semantic features," In CVPR Workshop (CVPRW) on the AI City Challenge, 2018
Hao Xiao, Weiyao Lin, Bin Sheng, Ke Lu, Junchi Yan, Jingdong Wang, Errui Ding, Yihao Zhang, and Hongkai Xiong. 2018. Group Re-Identification: Leveraging and Integrating Multi-Grain Information. In 2018 ACM Multimedia Conference
- Construction and Visualization of Heterogeneous Topic Web Based on Text Network, Patent #: ZL201610757401.0 (Chinese) [Link]
Student Travel Grant, ACM Multimedia 2018 conference, funded by US National Science Foundation(NSF), Aug. 2018
Winner Team, Track 1 & Track 3 at the NVIDIA AI City Challenge Workshop at CVPR 2018, Jun. 2018
Academic Excellence Scholarship & Xin Dong Scholarship & National Endeavor Fellowship in SJTU, 2013-2017
First Prize, Chinese Mathematical Olympiad & China Undergraduate Mathematical Contest in Modeling, 2012-2013
- Conference and Journal Reviewer: CVPR 2018, ETRI 2018