研究Research

用人工智能理解地球AI for understanding the Earth

我的研究位于地球物理、计算方法与机器学习的交叉地带，关注能够深化科学认识、并在真实场景中保持价值的方法。I work at the intersection of geophysics, computation, and machine learning, with an interest in methods that deepen scientific understanding and remain useful in real-world settings.

代表性工作 Selected Work

一些近期研究 A selection of recent research

FiLark streaming-first architecture connecting DAS exploration, analysis and annotation, and algorithm integration
01

FiLark: 面向分布式声学传感端到端探索、标注与算法集成的流式优先软件框架 FiLark: a streaming-first software framework for end-to-end exploration, annotation, and algorithm integration in distributed acoustic sensing

面向连续、大规模 DAS 数据的流式优先框架，将交互探索、信号处理、事件标注与算法集成连接为统一工作流。 A streaming-first framework that unifies interactive exploration, signal processing, event annotation, and algorithm integration for continuous, large-scale DAS data.

Li, J., Li, W.†, Tong, K., and Guo, X., 2026. FiLark: a streaming-first software framework for end-to-end exploration, annotation, and algorithm integration in distributed acoustic sensing. Manuscript under review. Li, J., Li, W.†, Tong, K., and Guo, X., 2026. FiLark: a streaming-first software framework for end-to-end exploration, annotation, and algorithm integration in distributed acoustic sensing. Manuscript under review.

  • 分布式声学传感 Distributed Acoustic Sensing
  • 流式数据处理 Streaming Data Processing
  • 科学软件 Scientific Software
PreprintDocs (Preview)
Deep Learning-based Seismic Reflectivity Estimation by Pre-training on Labeled Synthetic Data and Physics-guided Fine-tuning in Field Data
02

Deep Learning-based Seismic Reflectivity Estimation by Pre-training on Labeled Synthetic Data and Physics-guided Fine-tuning in Field Data Deep Learning-based Seismic Reflectivity Estimation by Pre-training on Labeled Synthetic Data and Physics-guided Fine-tuning in Field Data

Physics-guided fine-tuning for seismic reflectivity estimation from synthetic pre-training to field-data adaptation. Physics-guided fine-tuning for seismic reflectivity estimation from synthetic pre-training to field-data adaptation.

Wang, Y., Li, J.†, Sun, X., and Wu, X.†, 2026. Deep Learning-based Seismic Reflectivity Estimation by Pre-training on Labeled Synthetic Data and Physics-guided Fine-tuning in Field Data. Geophysical Journal International: ggag317. Wang, Y., Li, J.†, Sun, X., and Wu, X.†, 2026. Deep Learning-based Seismic Reflectivity Estimation by Pre-training on Labeled Synthetic Data and Physics-guided Fine-tuning in Field Data. Geophysical Journal International: ggag317.

  • Reflectivity Estimation Reflectivity Estimation
  • Physics-Guided Learning Physics-Guided Learning
  • Seismic Inversion Seismic Inversion
Paper
High‑Fidelity Seismic Super‑Resolution Using Prior‑Informed Deep Learning with 3D Awareness
03

High‑Fidelity Seismic Super‑Resolution Using Prior‑Informed Deep Learning with 3D Awareness High‑Fidelity Seismic Super‑Resolution Using Prior‑Informed Deep Learning with 3D Awareness

Prior-informed deep learning with 3D awareness for high-fidelity seismic super-resolution in realistic field data. Prior-informed deep learning with 3D awareness for high-fidelity seismic super-resolution in realistic field data.

Li, J., Wu, X., Zhang, X., Du, X., Sun, X., Deng, B. and Wang, G., 2025. High‑Fidelity Seismic Super‑Resolution Using Prior‑Informed Deep Learning with 3D Awareness. IEEE Transactions on Image Processing. Li, J., Wu, X., Zhang, X., Du, X., Sun, X., Deng, B. and Wang, G., 2025. High‑Fidelity Seismic Super‑Resolution Using Prior‑Informed Deep Learning with 3D Awareness. IEEE Transactions on Image Processing.

  • Seismic Super-Resolution Seismic Super-Resolution
  • Prior-Informed Learning Prior-Informed Learning
  • 3D Awareness 3D Awareness
PaperPDF
Memory-Efficient Full-Volume Inference for Large-Scale 3D Dense Prediction without Performance Degradation
04

Memory-Efficient Full-Volume Inference for Large-Scale 3D Dense Prediction without Performance Degradation Memory-Efficient Full-Volume Inference for Large-Scale 3D Dense Prediction without Performance Degradation

Retraining-free operator-level optimization for memory-efficient full-volume inference in large-scale 3D dense prediction. Retraining-free operator-level optimization for memory-efficient full-volume inference in large-scale 3D dense prediction.

Li, J. and Wu, X., 2026. Memory-Efficient Full-Volume Inference for Large-Scale 3D Dense Prediction without Performance Degradation. Communications Engineering.
(This paper was unanimously recognized by all three named reviewers: Peer Review File) Li, J. and Wu, X., 2026. Memory-Efficient Full-Volume Inference for Large-Scale 3D Dense Prediction without Performance Degradation. Communications Engineering.
(This paper was unanimously recognized by all three named reviewers: Peer Review File)

  • 3D Dense Prediction 3D Dense Prediction
  • Inference Efficiency Inference Efficiency
  • Operator Optimization Operator Optimization
PaperPDFCodeSupp
Multidimensional geophysical data visualized with CIGVis, including seismic volumes, horizons, faults, and geological bodies
05

CIGVis: an open-source Python tool for real-time interactive visualization of multidimensional geophysical data CIGVis: an open-source Python tool for real-time interactive visualization of multidimensional geophysical data

Open-source Python software for real-time interactive visualization of multidimensional geophysical data. Open-source Python software for real-time interactive visualization of multidimensional geophysical data.

Li, J., Shi, Y., and Wu, X., 2024. CIGVis: an open-source Python tool for real-time interactive visualization of multidimensional geophysical data. Geophysics 90, 1–37. Li, J., Shi, Y., and Wu, X., 2024. CIGVis: an open-source Python tool for real-time interactive visualization of multidimensional geophysical data. Geophysics 90, 1–37.

  • Visualization Visualization
  • Open-Source Software Open-Source Software
  • 3D Geophysical Data 3D Geophysical Data
PaperCodeDocsPyPI