Feng Wang(王峰)

Ph. D. student @ UESTC

I am a perception algorithm engineer at TuSimple. Before that, I got my Ph.D. degree from UESTC, Jian Cheng's Group. During my Ph.D. life, I was luckily mentored by Alan Yuille as a visiting student at Johns Hopkins University. My research interest includes Deep Learning and Facial Image Analysis.
Social links:   Github    Weibo     Zhihu     Blog
Academic links:     Google Scholar     Research Gate   

Fingerprint Liveness Detection

  • Designed a hand-crafted feature called Ridge-Slope-Valley feature to sample more information compared with previous works. Submitted to an EI indexed conference.
  • Concluded some statistical features into a new framework, proposed some techniques to promote the performance of the framework. Paper Link


Aircraft Docking System

  • System architecture construction, including submodule scheduling, message process, COM and TCPIP communication and parameter configuration.
  • Laser point cloud processing algorithm design and test. It includes calibration, coordinate transformation, active tracking and signal processing.
  • Machine Vision algorithm design, exploit and test. It includes moving object detection, airplane engine detection and tracking, airplane front wheel detection and tracking. 3 Chinese patents applied.
  • Video demonstration.

engineering system architectcore algorithm design

Open-source Software

  • The most starred unofficial windows version of Caffe.Star
  • C++ version of testing MTCNN. Star
  • C++ version of testing Faster RCNN. Star


Face Verification

  • Re-implementation of Joint Bayesian. [code]
  • NormFace, a new loss function, 99.21% on LFW. Star
  • Additive Margin Softmax, a new loss function. One improvement on NormFace. Star

engineering research

Deep Neural Network Visualization

  • Designed an algorithm to visualize how complex the neural network's fitting function can achieve, w.r.t. width, depth and structure.[code]
  • Improved the neuron visualization algorithm proposed by VGG team. [code1] and [code2]
  • Re-implement Neural Art. [code]
  • Re-implement Wasserstein GAN. [code]



LiDAR R-CNN: An Efficient and Universal 3D Object Detector
Zhichao Li*, Feng Wang*, Naiyan Wang.
CVPR 2021, Poster. [paper][code]

Additive Margin Softmax for Face Verification
Feng Wang, Weiyang Liu, Haijun Liu, Jian Cheng.
Signal Processing Letters, IEEE, 2018. ICLR 2018 workshop. [paper][code]

Visualizing Deep Neural Network by Alternately Image Blurring and Deblurring
Feng Wang, Haijun Liu, Jian Cheng.
Neural Networks, Elsvier, 2018. [paper][code][supplementary material]

NormFace: L2 HyperSphere Embedding for Face Verification
Feng Wang, Xiang Xiang, Jian Cheng, Alan L. Yuille.
ACM MM 2017, Poster. [paper][code][slides]

Regularizing Face Verification Nets For Pain and Expression Regression
Feng Wang, Xiang Xiang, Chang Liu, Trac D. Tran, Austin Reiter, Gregory D. Hager, Harry Quon, Jian Cheng, Alan L. Yuille.
ICIP 2017, Oral. [paper][code][slides]

Ridge-Slope-Valley Feature for Fingerprint Liveness Detection.
Feng Wang, Jian Cheng, and Yan Jiang.
CSPS 2015.[paper]

Research on Deep Learning Based Face Verification.
Ph.D. dissertation.[download][slides]

Non-first author papers:  see Google Scholar.


Ph. D. in Signal and Information Processing. Jian Cheng's Group.
University of Electronic Science and Technology of China

September 2014 �?now

Visiting Student in Alan Yuille's Group
Johns Hopkins University

November 2016 �?October 2017

MS in Signal and Information Processing. Jian Cheng's Group.
University of Electronic Science and Technology of China
third-class people's scholarship of year 2012
second-class people's scholarship of year 2013

September 2012 �?July 2014

BS in Electronic Information Engineering
University of Electronic Science and Technology of China

September 2008 �?July 2012



Created by BLACKTIE.CO