Understanding vision.
Exploring generation.理解视觉世界，
探索生成边界。
Focusing on Computer Vision, Generative AI, and Remote Sensing. Dedicated to exploring the boundaries of robust image translation and generation.
研究方向为计算机视觉、生成式人工智能与遥感，致力于探索鲁棒图像转换与生成的边界。
My current work investigates how generative models and robust visual representations can support trustworthy translation, synthesis, and interpretation of complex remote sensing imagery.
当前工作探究生成模型与鲁棒视觉表征如何支撑复杂遥感影像的可信转换、合成与解译。
News近期动态
- 2026
- 2026
- 2026σ: Sigmoid Modulation for Ultra High Resolution DiffusionAccepted at录用于 ICML
- 2026Batch Loss Score for Dynamic Data PruningAccepted at录用于 CVPRHighlight
- 2026Inconsistency Biases in Dynamic Data PruningAccepted at录用于 ICLR
- 2025
- 2025MMO-IG: Multi-Class and Multi-Scale Object Image Generation for Remote SensingAccepted at录用于 T-GRS
- 2025Edge Approximation Text DetectorAccepted at录用于 T-CSVT
Selected Publications代表性论文
View full figure查看完整配图StructDiff: A Training-Free Framework for Precise Structural Control in Remote Sensing Image Synthesis
A training-free framework for precise structural control in remote sensing image synthesis, combining multimodal prompt modulation, fixed-timestep structural conditioning, and restart refinement. Accepted by IEEE T-MM.
面向遥感影像合成的免训练精确结构控制框架，结合多模态提示调制、固定时间步结构条件约束与重启细化策略。已被 IEEE T-MM 接收。
Code代码
View full figure查看完整配图SHARP: Spectrum-aware Highly-dynamic Adaptation for Resolution Promotion in Remote Sensing Synthesis
Training-free large-scale remote sensing text-to-image synthesis with an RS-adapted FLUX prior and dynamic positional adaptation.
免训练的大尺度遥感文生图方法，结合面向遥感适配的 FLUX 先验与动态位置自适应。
Code代码
View full figure查看完整配图σ: Sigmoid Modulation for Ultra High Resolution Diffusion
A sigmoid-modulated framework for stable and efficient ultra high resolution diffusion generation.
基于 sigmoid 调制的框架，实现稳定高效的超高分辨率扩散生成。
Code代码
View full figure查看完整配图Batch Loss Score for Dynamic Data Pruning
A dynamic data pruning method based on batch loss scoring, accepted as a CVPR 2026 Highlight paper.
基于批次损失打分的动态数据剪枝方法，入选 CVPR 2026 Highlight。
Code代码
View full figure查看完整配图Inconsistency Biases in Dynamic Data Pruning
A study of inconsistency bias in dynamic data pruning, examining how resampling, local-window pruning, and cumulative temporal rescaling affect training dynamics.
研究动态数据剪枝中的不一致性偏差，考察重采样、局部窗口剪枝与累积时间重缩放对训练动态的影响。
Code代码
View full figure查看完整配图RLI-DM: Robust Layout-Based Iterative Diffusion Model for SAR-to-RGB Image Translation
A diffusion-based model for translating SAR images to optical (RGB) images, enhancing structural consistency.
基于扩散模型的 SAR 到光学（RGB）图像转换方法，显著提升结构一致性。
View full figure查看完整配图MMO-IG: Multi-Class and Multi-Scale Object Image Generation for Remote Sensing
A generative framework for synthesizing remote sensing objects across categories and spatial scales.
面向遥感的生成框架，可跨类别与跨空间尺度合成目标。
Code代码
View full figure查看完整配图Edge Approximation Text Detector
A text detection approach that improves localization by modeling contour structure and edge approximation.
通过建模轮廓结构与边缘逼近来提升定位精度的文本检测方法。
View full figure查看完整配图Traffic Sign Interpretation via Natural Language Description
An interpretable framework that connects traffic sign understanding with natural language generation.
将交通标志理解与自然语言生成相连接的可解释框架。
Education教育背景
Northwestern Polytechnical University西北工业大学
Ph.D. Candidate in Computer Science and Engineering
计算机科学与技术 博士研究生
Research focus: Computer Vision.
研究方向：计算机视觉。
Northwestern Polytechnical University西北工业大学
B.S. in Mathematics
数学 学士
Solid foundation in mathematics.
打下扎实的数学基础。
Honors荣誉奖项
- Outstanding Graduate Student (PhD)优秀研究生（博士）Northwestern Polytechnical University西北工业大学2025
- Second Prize Academic Scholarship (PhD)学业奖学金二等奖（博士）Northwestern Polytechnical University西北工业大学2025
- First Prize Academic Scholarship (PhD)学业奖学金一等奖（博士）Northwestern Polytechnical University西北工业大学2024
- First Prize Scholarship (Undergraduate)一等奖学金（本科）Northwestern Polytechnical University西北工业大学2023
- Second Prize Scholarship (Undergraduate)二等奖学金（本科）Northwestern Polytechnical University西北工业大学2022
Good research starts with a conversation.让好的研究，从一次交流开始。
If you are interested in collaboration or discussion, feel free to reach out via email.
如有合作或交流意向，欢迎通过邮件与我联系。