Weifeng Chen

I am a final-year Ph.D. student at the University of Michigan, Ann Arbor, advised by Prof. Jia Deng. Previously, I received my B. Eng degree from Zhejiang University under the supervision of Prof. Guofeng Zhang. I am currently visiting Jia's lab at Princeton University.

I work on recovering 3D properties from images with deep learning. I mainly explore ways to efficiently collect high-quality 3D annotations from the Internet to train 3D-perception networks.

Email  /  CV  /  Google Scholar  /  Github

Learning Single-Image Depth from Videos using Quality Assessment Networks
Weifeng Chen, Shengyi Qian, Jia Deng
Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
[paper][project site][code]

We propose a method to automatically generate training data for single-view depth through Structure-from-Motion (SfM) on Internet videos.

Surface Normals in the Wild
Weifeng Chen, Donglai Xiang, Jia Deng
International Conference on Computer Vision (ICCV), 2017.
Invited [poster] at the Bridges to 3D Workshop, CVPR 2018

We present a new dataset "Surface Normals in the Wild" consisting of images in the wild annotated with surface normals of random points.

Single-Image Depth Perception in the Wild
Weifeng Chen, Zhao Fu, Dawei Yang, Jia Deng
Neural Information Processing Systems (NeurIPS), 2016.
[paper][dataset][code][supplementary material][BibTex]

We introduce a new dataset "Depth in the Wild" consisting of images in the wild annotated with relative depth between pairs of random points.

Featured in the Wolfram Neural Net Repository. See this article for more details.
Multi-Viewpoint Panorama Construction with Wide-Baseline Images
Guofeng Zhang, Yi He, Weifeng Chen, Jiaya Jia and Hujun Bao
IEEE Transactions on Image Processing (TIP), 2016.

We design a mesh-based framework for creating panoramas from wide-baseline images.

High-Quality Depth Recovery via Interactive Multi-View Stereo
Weifeng Chen, Guofeng Zhang, Xiaojun Xiang, Jiaya Jia and Hujun Bao
International Conference on 3D Vision (3DV), 2014.

We align CAD models interactively to fix artifacts in MVS output.

Learning to Sit: Synthesizing Human-Chair Interactions via Hierarchical Control
Yu-Wei Chao, Jimei Yang, Weifeng Chen, Jia Deng

We propose a learning framework that learns to synthesize realistic human motion of sitting.


Program Committee (PC): AAAI 2020
Reviewer: ICCV 2019, ICML 2019, CVPR 2019 & 2020, NeurIPS 2018 & 2019, SIGGRAPH Asia 2018, WACV 2020

Template Credit: Jon Barron