Faculty Profile

Xi Peng

Xi Peng

Assistant Professor
440 Smith Hall
Newark, Delaware 19716
P: 302-831-2876

Personal Website


PhD | 2018 | Rutgers, The State University of New Jersey
MS | 2011 | Chinese Academy of Sciences, Beijing, China
BS | 2008 | Beihang University, Beijing, China


Dr. Xi Peng is an assistant professor of the Department of Computer & Information Sciences, and a resident faculty of the Data Science Institute, at the University of Delaware.

Dr. Peng works in the area of Deep Learning, Machine Learning, and Computer Vision, with a special interest in two directions: 1) Efficient & explainable deep learning; 2) Human-centered computer vision. He is making efforts in developing strong, flexible, and transparent AI systems for interdisciplinary data analytics in biomechanics, bioinformatics, geography, and earth science.

Dr. Peng’s group (Deep-REAL lab) are publishing in top-ranked conferences (NeurIPS, ICLR, CVPR, ICCV, ECCV, AAAI, IJCAI, KDD), leading journals (TPAMI, IJCV, THMS), as well as US patents. His group received awards and supports from NSF, CDC/NIOSH, Snap Research, DSI, etc.


  1. [AAAI’21] Multimodal learning with severely missing modality.
  2. [ICLR’21] A good image generator is what you need for high-resolution video synthesis.
  3. [CVPR’20] Learning to learn single domain generalization.
  4. [CVPR’20] Knowledge as priors: cross-modal knowledge generalization.
  5. [NeurIPS’20] Maximum-entropy adversarial data augmentation for generalization and robustness.
  6. [NeurIPS’19] Rethinking kernel methods for node representation learning on graphs.
  7. [NeurIPS’19] Semantic-guided multi-attention localization for zero-shot learning.
  8. [CVPR’19] Semantic graph convolutional networks for 3d human pose regression.
  9. [KDD’19] Scalable global alignment graph kernel using random features.
  10. [ICCV’19] Adaptive data transformation.
  11. [TPAMI’19] Towards efficient u-nets: a coupled and quantized approach.