- B.S. in Computer Science, Peking University, 2021 (expected)
- Peking University, Academic Excellence Award, 2020
- CVPRW NTIRE Challenge Image Demoireing Burst Track, Winner, 2020
- CVPRW NTIRE Challenge Image Demoireing Single Image Track, Runner-up Award, 2020
- CVPRW NTIRE Challenge Image Deblurring Desktop Track, Runner-up Award, 2020
- National Olympiad in Informatics, Bronze medal, 2015, 2016
- National Olympiad in Informatics Winter Camp, Silver medal, 2016
- 2020.2 - now: Research Intern
- CyLab, Electrical and Computer Engineering, Carnegie Mellon University
- Advised by Dr. Zhiqiang Shen and Prof. Marios Savvides.
- Knowledge Distillation: Clarifying a recently discovered perspective that label smoothing is incompatible with knowledge distillation.
- 2018.7 - 2019.12: Research Intern
- STRUCT Team, Wangxuan Institute of Computer Technology, Peking University
- Advised by Prof.~Jiaying Liu.
- Single Image Deraining: Proposed Fractal Band Learning with self-supervision strategy, improved PSNR by 1.74dB on Rain100L and by 0.64dB on Rain800 compared to baseline (RESCAN, RWL). [Github].
- Joint Inpainting and Colorization: Adopted Gated Convolution based UNet which effectively restores old photos. [Github].
- Low-light Enhancement and Face Detection: Proposed a new benchmark for low-light enhancement and a SOTA model (+2.1 mAP) for face detection in low-light environment. [Project].
- 2020.1 - now: Megvii Technology Inc
- Research Intern. Advised by Jiangyu Liu.
- Eyeglass Reflection Removal: Refined occlusion segmentation module and error protection classifier for the eyeglass reflection removal pipeline which landed in recent smartphones including NOVA 7.
- 2020.6 - 2020.7: DeeCamp, Sinovation Ventures Inc.
- Deep learning based program repair product for online programming course.
- 2020.6 - 2020.8: Tensorflow, Google Summer of Code.
- Intern, Mentored by Jaeyoun Kim and Mandar Deshpande.
- Implement Generative Inpainting (ICCV’19) and Learning to see in the dark (CVPR’18) with Tensorflow 2 for Tensorflow Model Garden.
- Dejia Xu, Wenhan Yang, Minhao Fan, Jiaying Liu. “Benchmarking Low-Light Image Enhancement and Beyond”, IJCV, to appear in 2021.
- Wenhan Yang, Shiqi Wang, Dejia Xu, Xiaodong Wang and Jiaying Liu. “Towards Scale-Free Rain Streak Removal via Self-Supervised Fractal Band Learning,” Proc. AAAI Conf. on Artificial Intelligence, 2020.
- Dejia Xu, Yihao Chu, Qingyan Sun. “Moire Pattern Removal via Attentive Fractal Network Pattern Removal via Attentive Fractal Network”, CVPRW, 2020.
- “NTIRE 2020 Challenge on Image Demoireing: Methods and Results”, CVPRW, 2020.
- “NTIRE 2020 Challenge on Image and Video Deblurring”, CVPRW, 2020.
- “AIM 2020 challenge on real image super-resolution: Methods and results”, ECCVW, 2020.
- “AIM 2020 challenge on image extreme inpainting”, ECCVW, 2020.
- Zhiqiang Shen, Zechun Liu, Dejia Xu, Zitian Chen, Kwang-Ting Cheng, Marios Savvides. “Is Label Smoothing Truly Incompatible with Knowledge Distillation: An Empirical Study”, submitted to ICLR, 2021.
- Dejia Xu, Haoning Wu, Yu Han. “Stereo Image Super Resolution via StereoScopic Guidance Network”, submitted to CVPR, 2021.
Service and leadership
- 2020.7 - 2020.9: Mentor
- Open Source Promotion Plan, Institute of Software, Chinese Academy of Sciences
- Mentored the development of the toolchain which exports markdown to LaTeX and lints customized markdown syntax. [Github].
- Guided the development and refinement of documentation. [Github].
- 2018.7 - 2019.9: Organizer
- UG2+ Challenge, CVPR 2019
- Co-organized the workshop and challenge: “Bridging the Gap between Computational Photography and Visual Recognition (UG2+)”. [Website].
- 2018.4 - now: Chair
- PKU Student Association of Algorithms
- Maintained the OI Wiki project and its ecosystem, which aims to share knowledge about competitive programming. [Github].
- Co-organized the 2018 ICPC World Final and ICPC Asia Beijing Regional. [Website].