Xinge Yang

I am a Computer Science PhD student in VCC Computational Imaging Group at KAUST, working with Prof. Wolfgang Heidrich.

My research interests lie on co-design of optics and computer vision and optics-based computer vision. We created a differentiable ray tracer DeepLens that can simulate camera sensor image by ray-tracing rendering, and directly optimize sensor image for optical design.

I am also a marathon runner(personal record: half-marathon - 79mins, marathon - 3h01min), basketball player, and photographer.

Email me to start a talk. I am open to 2024 internship position. :-)

Email  /  CV  /  Github  /  LinkedIn  /  ηŸ₯乎

profile photo
Education

  • 2022 - present: Ph.D. student in Computer Science, KAUST, Saudi Arabia.
  • 2020 - 2022: M.Sc. in Computer Science, KAUST, Saudi Arabia.
  • 2016 - 2020: B.Sc. in Physics (major) and Computer Science (minor), USTC, China.

    		•
    News

  • 07/2023: Our paper, "Aberration-Aware Depth-from-Focus" is accepted by ICCP and TPAMI, let's have a coffee in Madison!
  • 05/2023: Our paper "Image Quality Is Not All You Want: Task-Driven Lens Design for Image Classification" is avaliable on Arxiv. paper link.
  • 05/2023: I will attend Siggraph this year, see you in LA!
  • 02/2023: The DeepLens paper, "Curriculum Learning for ab initio Deep Learned Refractive Optics" is avaliable on Arxiv. paper link.
  • 01/2023: Happy new year! Please have a look at "awesome-deep-optics" repo if you are intested in optics and network co-design.
  • 05/2022: Our paper "Automatic Lens Design based on Differentiable Ray-tracing" is accepted by OSA Imaging Congress - COSI, see you in Vancouver!
  • 04/2022: I got the opportunity from my advisor to attend Siggraph this year, see you in Vancouvar!
  • 03/2022: I defended my Master thesis "Automatic Lens Design based on Differentiable Ray-tracing"!
  • 09/2020: I started Master program at KAUST!

    		•
    Research

    I do co-design of optics and computer vision, with a focus on differentiable optical simulation. I am maintaining a differentiable ray tracer ("DeepLens") in our group.

    		Image Quality Is Not All You Want: Task-Driven Lens Design for Image Classification
    Xinge Yang, Qiang Fu, Yunfeng Nie, Wolfgang Heidrich
    Arxiv prepint. paper

  • Use a network to supervise lens design from scartch.
  • Our image classification lens can achieve higher accuracy with fewer elements compared to conventional lenses.

    		•
    		Aberration-Aware Depth-from-Focus
    Xinge Yang, Qiang Fu, Mohamed Elhoseiny, Wolfgang Heidrich
    ICCP & TPAMI 2023. paper, code (coming soon)

  • Use a network to represent optical aberrations in real camera lenses.
  • Improve the generalizability of network models by considering optical aberrations during the training.

    		•
    		Curriculum Learning for ab initio Deep Learned Refractive Optics
    Xinge Yang, Qiang Fu, Wolfgang Heidrich
    Arxiv prepint. paper,   code (coming soon),   video
    OSA Imaging and Applid Optics Congress - Computational Optical Sensing and Imaging (COSI) 2022 (oral): paper,   proj

  • Make lens design as similar and easy as network training.
  • Automatically design both classical and computational lenses from scratch.


    • The website template is from Dr. John Barron.