I am a senior research scientist at Google, working closely with Cordelia Schmid and Kevin Murphy. I received my Ph.D. from University of Southern California in 2015, advised by Prof. Ram Nevatia. During graduate school, I was lucky enough to work with Sanketh Shetty and Rahul Sukthankar at Google Research; Lubomir Bourdev, Ronan Collobert and Manohar Paluri at Facebook AI Research. Before coming to the U.S., I received my B.Eng. in Computer Science at Tsinghua University in 2011. And even before that, I attended Yaohua High School in Tianjin, China.
My current research interests include human action recognition and dynamics prediction from videos. I used to work on object detection and webly-supervised learning. Together with my amazing colleagues, I won the COCO object detection challenge 2016 and iNaturalist challenge 2017. Our object detection algorithms have been open sourced as the Tensorflow Object Detection API. To facilate research on machine perception, I also work on dataset collection, notably the Atomic Visual Actions dataset for human action recognition, the Open Images dataset for object detection and the iNaturalist dataset for fine-grained recognition.
Here are a few representative recent publications:
Human Action Recognition
Saining Xie, Chen Sun, Jonathan Huang, Zhuowen Tu and Kevin Murphy. Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification. ECCV 2018. [arXiv] [mini-Kinetics splits] [Code (coming soon)]
- Chen Sun, Per Karlsson, Jiajun Wu, Joshua B. Tenenbaum, and Kevin Murphy, Predicting the Present and Future States of Multi-agent Systems from Partially-observed Visual Data. ICLR 2019.
- Chen Sun, Abhinav Shrivastava, Saurabh Singh, and Abhinav Gupta, Revisiting Unreasonable Effectiveness of Data in Deep Learning Era. ICCV 2017 (spotlight). [arXiv] [Research Blog] [Wired]
For a full list please refer to my Google Scholar page.