Qi Li

Qi Li

Senior Undergraduate

Tsinghua University

Biography

I am a senior undergraduate in MARS Lab at Tsinghua University, supervised by Prof. Hang Zhao. Currently I am taking an internship at Stanford University, advised by Prof. Leonidas J. Guibas. I also had memoriable internships at Microsoft Research Asia and Kuaishou.

My research interests are in autonomous driving, robotics, computer vision and AutoML. Currently I am looking for a 2022 Fall Ph.D. position.

Download my resumé.

Interests
  • Autonomous Driving
  • Robotics
  • Computer Vision
  • AutoML
Education

  • BSc in Software Engineering, 2018

    Tsinghua University

Publications

IFR-Explore: Learning Inter-object Functional Relationships in 3D Indoor Scenes
We take the first step in building AI system learning inter-object functional relationships in 3D indoor environments with key technical contributions of modeling prior knowledge by training over large-scale scenes and designing interactive policies for effectively exploring the training scenes and quickly adapting to novel test scenes.
Qi Li, Kaichun Mo, Yanchao Yang, Hang Zhao, Leonidas Guibas
HDMapNet: An Online HD Map Construction and Evaluation Framework
we introduced a local semantic map learning method, dubbed HDMapNet, which encodes image features from surrounding cameras and/or point clouds from LiDAR, and predicts vectorized map elements in the bird’s-eye view.
Qi Li, Yue Wang, Yilun Wang, Hang Zhao
PDF Cite Code Video Project Page
Cream of the Crop: Distilling Prioritized Paths For One-Shot Neural Architecture Search
We introduce the concept of prioritized path, which refers to the architecture candidates exhibiting superior performance during training. Distilling knowledge from the prioritized paths is able to boost the training of subnetworks.
Houwen Peng, Hao Du, Hongyuan Yu, Qi Li, Jing Liao, Jianlong Fu
PDF Cite Code
Negative Margin Matters: Understanding Margin in Few-shot Classification
We introduced a negative margin loss to metric learning based few-shot learning methods. The negative margin loss significantly outperforms regular softmax loss, and achieves state-of-the-art accuracy on three standard few-shot classification benchmarks with few bells and whistles. These results are contrary to the common practice in the metric learning field, that the margin is zero or positive.
Bin Liu, Yue Cao, Yutong Lin, Qi Li, Zheng Zhang, Mingsheng Long, Han Hu
PDF Cite Code

Experience

 
 
 
 
 
Stanford, Advisor: Leonidas Guibas
Research Assistant (UGVR program)
Stanford, Advisor: Leonidas Guibas
Mar 2021 – Present Beijing (remote)
Conducted research on a new benchmark of learning indoor multi-object functional relationships.
 
 
 
 
 
Tsinghua, Advisor: Hang Zhao
Research Assistant
Tsinghua, Advisor: Hang Zhao
Dec 2020 – Present Beijing
Conducted research on a new benchmark for predicting local HD Map using onboard sensors.
 
 
 
 
 
Microsoft Research Asia, Advisor: Houwen Peng
Research Assistant
Microsoft Research Asia, Advisor: Houwen Peng
Jan 2020 – Sep 2020 Beijing
Conducted research on Neural Architecture Search.
 
 
 
 
 
Tsinghua, Advisor: Mingsheng Long
Research Assistant
Tsinghua, Advisor: Mingsheng Long
Aug 2019 – Nov 2019 Beijing
Integrated negative margin to the softmax loss to benefit the few-shot classification.
 
 
 
 
 
MMU, KuaiShou
Engineer Intern
MMU, KuaiShou
Jun 2019 – Aug 2019 Beijing
Detected offensive content in videos acquired from the Kuaishou App using synthetic model.

Contact