I am a Ph.D. Candidate at Fei Xia Research Lab @ UC, Irvine, with research interests in Computational Imaging & Photography, particularly in imaging through scattering, image restoration and generative models. I have worked in ByteDance, OPPO and VIVO on computational photography and ISP algorithms.
🔥 News
- 2026.11, Looking for 2027 Summer Internships.
- 2026.06, Joining Bytedance(PICO) as Research Scientist Intern!
- 2026.04, One of our papers has been accepted by the journal Advanced Photonics.
📖 Educations
- 2025.09 - Present, Ph.D Candidate, EECS, UC Irvine
. Ph.D Supervisor: Prof. Fei Xia - 2021.09 - 2024.06, Master, EE, Xi’an Jiaotong Univerisity
. Master Supervisor: Prof. Jianbin Liu - 2017.09 - 2021.06, Bachelor, EEE, Huazhong University of Science and Technology

💻 Experience
- 2026.06 - 2026.09, Research Scientist Intern, PICO-Lab, Bytedance
, San Jose, United States. - 2024.07 - 2025.04, Machine Learning Engineer, YLab, OPPO Research Institute
, China. Advisor: Prof. Lei Zhang - 2023.03 - 2023.05, Machine Learning Intern, VIVO Dept. of Image Effect
, China.
🔠Research
🔬 Computational Imaging

NeuSPI: Neural Single-Pixel Imaging Turning Motion into Information with Sub-Pixel Precision
Y Sun, J Seo and F Xia.
- We introduce NeuSPI, a space-time joint reconstruction algorithm with neural representation. It works with motion-resilient imaging, dynamic imaging and pixel super-resolution with motion estimation.

NeOTF: Guidestar-free neural representation for broadband dynamic imaging through scattering
Y Sun and F Xia.
- We introduce NeOTF, a framework that learns an implicit neural representation of the OTF. By applying multi-frame speckle intensities and Fourier-domain prior, NeOTF robustly retrieves the system’s OTF with high fidelity and efficiency.
📷 Computational Photography

CRAW: Edge-Efficient Joint RAW Compression and Restoration for Smart-Wearable Devices
Y Sun, S Chen and F Xia.
- We proposed an edge-efficient joint RAW restoration and compression for smart-glasses computational photography with distributed compute platform. Efficiency and hardware profiling are evaluated on Snapdragon 8Gen2 and AR1 Gen1 devices.

Open-AISP: An open-source AI-ISP pipeline framework for beginners
[Github] [Docs] [䏿–‡æ–‡æ¡£]
- A beginner-friendly and toy-level AI-ISP pipeline framework covering realistic raw image degradation simulation, neural ISP pipeline reconstruction, and image post-enhancement with generative prior. Current modules include raw simulation, joint denoising/demosaicing (JDD), with HDR-fusion, AITM, and DiffIPE on the roadmap.