Zehao Huang

Currently, I am taking a short break to explore new opportunities.

From 2017 to 2024, I worked at TuSimple in Beijing, leading a small team focused on solving real-world perception problems in truck autonomous driving.

Email  /  GitHub  /  Google Scholar  

profile photo

Work Experience:

TuSimple: Tech Lead (2017.10 - 2024.03)

:
  • L4 Image Perception: image based 2D and 3D object perception, map element segmentation and detection
  • L2 Perception: BEV based 3D object and lane perception with Camera-LiDAR Fusion
  • Miscellaneous: neural network acceleration, and consistency-based data mining

Research Interests:

  • Perception of 2D and 3D Objects and Map Elements in Autonomous Driving.
  • Deep Network Acceleration and Auto ML.

Selected Publications:

(* Interns, + Equal contribution)
project image

SparseFusion: Efficient Sparse Multi-Modal Fusion Framework for Long-Range 3D Perception


Yiheng Li*, Hongyang Li, Zehao Huang, Hong Chang, Naiyan Wang
arXiv, 2024
arxiv

A simple sparse fusion framework for long-range 3D perception.
project image

TSTTC: A Large-Scale Dataset for Time-to-Contact Estimation in Driving Scenarios


Yuheng Shi*, Zehao Huang, Yan Yan, Naiyan Wang, Xiaojie Guo
arXiv, 2023
arxiv / code / website

We construct a dataset for Time-to-Concact (TTC) Estimation in driving scenarios and provide several baseline methods.
project image

Object as Query: Lifting any 2D Object Detector to 3D Detection


Zitian Wang*, Zehao Huang, Jiahui Fu, Naiyan Wang, Si Liu
ICCV, 2023
arxiv / code

We design Multi-View 2D Objects guided 3D Object Detector (MV2D), which can lift any 2D object detector to multi-view 3D object detection.
project image

Anchor3dlane: Learning to regress 3d anchors for monocular 3d lane detection


Shaofei Huang*, Zhenwei Shen, Zehao Huang, Zi-han Ding, Jiao Dai, Jizhong Han, Naiyan Wang, Si Liu
CVPR, 2023
arxiv / code

We define 3D lane anchors in the 3D space and propose a BEV-free method named Anchor3DLane to predict 3D lanes directly from perspective view representations.
project image

QueryDet: Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection


Chenhongyi Yang*, Zehao Huang, Naiyan Wang
CVPR (Oral), 2022
arxiv / code

We propose QueryDet that uses a novel query mechanism to accelerate the inference speed of feature-pyramid based object detectors
project image

Direct Differentiable Augmentation Search


Aoming Liu*, Zehao Huang, Zhiwu Huang, Naiyan Wang
ICCV, 2021
arxiv / code

We propose an efficient differentiable search algorithm called Direct Differentiable Augmentation Search (DDAS). It exploits meta-learning with one-step gradient update and continuous relaxation to the expected training loss for efficient search.
project image

Learnable Graph Matching: Incorporating Graph Partitioning with Deep Feature Learning for Multiple Object Tracking


Jiawei He*, Zehao Huang, Naiyan Wang, Zhaoxiang Zhang
CVPR, 2021
arxiv / code

We propose a learnable graph matching method for object tracking.
project image

You Only Search Once: Single Shot Neural Architecture Search via Direct Sparse Optimization


Xinbang Zhang*, Zehao Huang, Naiyan Wang, Shiming Xiang, Chunhong Pan
TPAMI, 2021
arxiv / code

We propose a model pruning formulation for Neural Architecture Search (NAS) based on sparse optimization.
project image

1st Place Solutions of Waymo Open Dataset Challenge 2020 -- 2D Object Detection Track


Zehao Huang, Zehui Chen+, Qiaofei Li+, Hongkai Zhang, Naiyan Wang
arxiv, 2020
arxiv

1st place solution of Waymo Open Dataset Challenge 2020 in 2D detection track.
project image

Data-Driven Sparse Structure Selection for Deep Neural Networks


Zehao Huang, Naiyan Wang
ECCV, 2018
arxiv / code

We propose a framework to learn and prune deep models in an end-to-end manner.
project image

Like What You Like: Knowledge Distill via Neuron Selectivity Transfer


Zehao Huang, Naiyan Wang
arxiv, 2017
arxiv / code

We propose a framework for knowledge distillation by minimizing the Maximum Mean Discrepancy.




Design and source code from Leonid Keselman's website