Teng Xi was a joint postdoc researcher in Baidu Inc and Department of Computer Science and Technology at Tsinghua University. He received his Ph.D. degree from Beijing University of Posts and Telecommunications in 2018. He have been awarded the Chinese Government Scholarship for studying abroad, the National Scholarship for Ph.D. students, the BUPT Excellent Ph.D. Students Foundation. His research interest is model compression, Bayesian estimation, and face recognition. He has proposed original NAS algorithms in top-tier conferences/journals such as SA-NAS and GP-NAS. Using the proposed NAS algorithms, he received the Winner prize in all three tracks of the AIM 2020 Real Image Super-Resolution Challenge, the Winner prize in OVIC Image Track of 2020 Low-Power Computer Vision Challenge, the Winner prize in Grand Challenge of 106-Point Facial Landmark Localization, the Winner prize in NTIRE 2020 Challenge on Real Image Denoising. He co-organized the Astar 2019 Developer Challenge, which attracted nearly 2000 teams to attend, based on opensource PaddleSlim library developed by his group, a deep learning framework for model compression and architecture search. He co-organized the ICML Expo workshop PaddlePaddle-based Deep Learning at Baidu. He organized the CVPR 2021 Neural Architecture Search Workshop and organized the 1st lightweight NAS challenge. He organized the CVPR 2022 Neural Architecture Search Workshop and organized the 2nd lightweight NAS challenge.
■ VIMER-UFO 2.0 the largest CV model in the industry has been released! 05/20/2022
Introducing VIMER-UFO 2.0 (Unified Feature Optimization), a task MoE based 17-billion-parameter CV foundation model , which supports extraction of lightweight models through sparse activation and achieves SOTA on 28 datasets across a battery of visual recognition tasks including faces, humans, vehicles, commodities, food, etc. GitHub:https://github.com/PaddlePaddle/VIMER/tree/main/UFO/
■ Winner teams of CVPR 2022 NAS competition has been released! 06/07/2022