About me

I am an incoming master student in Robotics Institue at Carnegie Mellon University. Previously, I finished my undergraduate in School of EECS at Peking University, majoring in Computer Science advised by Prof. Zhanxing Zhu. I spent two years of my undergraduate life at Megvii Inc.(Face++) as a research intern mentored by Dr. Xiangyu Zhang. Later I joined Microsoft Research Asia as a research intern.

Research topics

  • Visual recognitaion.
  • Visual content creation/editing.
  • Autonomous Driving.
  • Understanding the prior of current deep learning models.

News


[November, 2020] We developed the first real-time flow based video frame interpolation methods with sota performance, named RIFE, see our demo here: RIFE.

[September, 2019] YOPO was accepted by NeurIPS19.

[July, 2019] One paper accepted by ICCV19.

[April, 2019] One paper accepted by ICML19.

[April. 2019] We published a new general detection dataset Objects365, which is designed to spur object detection research with a focus on diverse objects in the Wild.

Publications


RIFE: Real-Time Intermediate Flow Estimation for Video Frame Interpolation, Zhewei Huang, Tianyuan Zhang, Wen Heng, Boxin Shi, Shuchang Zhou. [project page], [code] arXiv preprint

Domain-Aware Dynamic Networks, Tianyuan Zhang, Bichen Wu, Xin Wang, Joseph Gonzalez, Kurt Keutzer. arXiv preprint

Objects365: A Large-scale, High-quality Dataset for Object Detection, Shuai Shao*, Zeming Li*, Tianyuan Zhang*, Chao Peng*, Gang Yu, Xiangyu Zhang, Jing Li, Jian Sun. (*equal contribution) ICCV, 2019

You Only Propagate Once: Accelerating Adversarial Training via Maximal Principle. Dinghuai Zhang*, Tianyuan Zhang*,Yiping Lu*, Zhanxing Zhu, Bin Dong. (*equal contribution) [code] 33rd Annual Conference on Neural Information Processing Systems (NeurIPS 2019)

Interpreting Adversarially Trained Convolutional Neural Networks. Tianyuan Zhang, Zhanxing Zhu. [code] International Conference on Machine Learning (ICML), 2019.