Buyu Deng

My name is Buyu Deng (邓卜瑜). I am currently a third-year undergraduate student majoring in Geophysics at the School of the Gifted Young, University of Science and Technology of China. Under the mentorship of Dr. Xinming Wu, I am conducting research on the application of deep learning techniques in geophysics.


Research Interests

Intelligent Processing of 3D Geophysical Data

3D geophysical data (e.g., seismic volumes) are characterized by their massive scale, low signal-to-noise ratios, and highly complex features. These properties pose significant bottlenecks for traditional data processing and interpretation. My research focuses on exploring and applying robust deep learning models (such as 3D-CNNs, Transformers, and Generative Models) to enable high-precision, automated analysis of these data volumes. The goal is to more accurately identify subsurface structures and invert physical parameters, overcoming long-standing challenges in applying effective AI to large-scale 3D geophysical datasets.

Insightful Applications of Cutting-Edge AI in Geophysics

I am interested in creatively adapting cutting-edge AI methodologies to solve fundamental geophysical problems. This includes, but is not limited to: applying advanced generative models (e.g., VAEs, GANs, Diffusion Models) for subsurface model generation and geophysical data reconstruction; developing multi-modal frameworks to fuse disparate datasets (such as seismic, well-log, and geological data); leveraging Physics-Informed Neural Networks (PINNs) for solving complex forward and inverse problems; and ensuring model robustness and interpretability through Explainable AI (XAI). I am committed to unlocking the potential of AI to provide novel, insightful perspectives for Geoscience.


Research Experience

  • 3D "All-in-One" Seismic Denoising

    • Constructed a high-fidelity synthetic 3D seismic dataset incorporating a comprehensive suite of common noise types (random, acquisition footprints, migration artifacts).
    • Trained an "All-in-One" denoising network based on Degradation Representation Learning, leveraging this specific, custom-built synthetic dataset to achieve robust noise suppression across all targeted degradation types.
    • Full code implementation available at: https://github.com/BuyuDeng/NDR_Seismic
  • 3D VAE for Geophysical Data Compression and Fusion

    • Engineered a novel 3D VAE by modifying the architecture of a pre-trained video model and fine-tuning it to suit the unique structural properties of 3D geophysical data.
    • This work resulted in a unified latent space representation for multi-source data fusion, enabling downstream tasks.
    • The findings are currently in preparation for publication.

News

  • August 2024   I joined Professor Xinming Wu’s research group, the Computational Interpretation Group (CIG), to engage in basic scientific training.
  • December 2024   My "Xinhe Researcher Program" ("Artificial Intelligence-Based Seismic Data Denoising") received an excellent evaluation during the proposal review.
  • January 2025   My "Undergraduate Research Program" ("Seismic Data Compression and Representation: A Deep Feature Representation Learning Framework Based on 3D Variational Autoencoders") was designated as a key project.
  • August 2025   I Participated in an intensive field program conducting geophysical surveys and engaged in academic seminars with faculty from the Institut de Physique du Globe de Paris (IPGP).

Awards

  • 2023 Freshman Scholarship, USTC
  • 2024 Qiangwei Scholarship (Endeavor), School of the Gifted Young, USTC
  • 2024 Zhao Jiuzhang Elite Program Scholarship (Category A), School of Earth and Space Sciences , USTC
  • 2025 Zhao Jiuzhang Elite Program Scholarship (Category B), School of Earth and Space Sciences , USTC

Publications