📝 Selected Publications

Annotation-Free Open-Vocabulary Segmentation for Remote-Sensing Images
Kaiyu Li, Xiangyong Cao, Ruixun Liu, Shihong Wang, Zixuan Jiang, Zhi Wang, Deyu Meng

Code

• We extend SegEarth-OV to SAR images. This is the first open vocabulary semantic segmentation work for SAR images.

SegEarth-R1: Geospatial Pixel Reasoning via Large Language Model
Kaiyu Li, Zepeng Xin, Li Pang, Chao Pang, Yupeng Deng, Jing Yao, Guisong Xia, Deyu Meng, Zhi Wang, Xiangyong Cao

Project | Dataset | Code

• We introduce the geospatial pixel reasoning task, construct the first benchmark dataset (EarthReason), and propose a simple yet effective baseline (SegEarth-R1).

Media coverage: 【遥感与深度学习】SegEarth-R1: 遥感+推理大模型！Geospatial Pixel Reasoning|【视觉与遥感前沿】SegEarth-R1实现地理推理「类人思考」|【减论】|

SegEarth-OV: Towards Training-Free Open-Vocabulary Segmentation for Remote Sensing Images
Kaiyu Li, Ruixun Liu, Xiangyong Cao, Xueru Bai, Feng Zhou, Deyu Meng, Zhi Wang

Project | Code | Bilibili | Youtube | Demo

• SegEarth-OV is the first to introduce training-free Open Vocabulary Semantic Segmentation into remote sensing images, which makes OVSS possible in remote sensing contexts.

Media coverage: 【遥感与深度学习】SegEarth-OV: 面向遥感图像的无训练开放词汇分割|【GISer阿兴】|【减论】|

DynamicEarth: How Far are We from Open-Vocabulary Change Detection?
Kaiyu Li, Xiangyong Cao, Yupeng Deng, Chao Pang, Zepeng Xin, Deyu Meng, Zhi Wang

Project

• We propose a new task, open-vocabulary change detection (OVCD).
• We propose two universal frameworks for OVCD.
• The first OVCD codebase, DynamicEarth, is released.

Open-CD: A Comprehensive Toolbox for Change Detection
Kaiyu Li, Jiawei Jiang, Andrea Codegoni, Chengxi Han, Yupeng Deng, Keyan Chen, Zhuo Zheng, Hao Chen, Zhengxia Zou, Zhenwei Shi, Sheng Fang, Deyu Meng, Zhi Wang, Xiangyong Cao

Project

• Open-CD is one of the most popular change detection toolkits.
• We launch the Open-CD Technical Report Plan (Open-CD TRP for shot). We invite some authors to introduce their algorithms and participate in the construction of the Open-CD codebase. This plan is under active development and we will keep this report updated.

SemiCD-VL: Visual-Language Model Guidance Makes Better Semi-supervised Change Detector
Kaiyu Li, Xiangyong Cao, Yupeng Deng, Junmin Liu, Deyu Meng, Zhi Wang

Code

• This work is the first to introduce visual language models to the semi-supervised change detection task.

A New Learning Paradigm for Foundation Model-based Remote Sensing Change Detection
Kaiyu Li, Xiangyong Cao, Deyu Meng

🏆️ ESI Highly Cited Paper

Code

• BAN is the first universal framework to adapt the foundation model to the change detection task.

Designing Domain-Specific Agents via Hierarchical Task Abstraction Mechanism
Kaiyu Li*, Jiayu Wang*, Zhi Wang, Hui Qiao, Weizhan Zhang, Deyu Meng, Xiangyong Cao

Project | GeoPlan-bench | EarthAgent Code

• We propose HTAM for designing domain-specific agents, instantiate this framework in the remote sensing field as EarthAgent, and introduce a comprehensive evaluation platform, GeoPlan-bench, to assess complex remote sensing planning capabilities.

DescribeEarth: Describe Anything for Remote Sensing Images
Kaiyu Li*, Zixuan Jiang*, Xiangyong Cao, Jiayu Wang, Yuchen Xiao, Deyu Meng, Zhi Wang

Dataset | Benchmark | Code

• We introduce geo-spatial detailed localized captioning.
• We build the first describe anything model in remote sensing.
• We release the related dataset and benchmark.

Media coverage: 【遥感与深度学习】DescribeEarth: 西交团队提出遥感描述一切模型, 鼠标轻点即得目标详细描述.|【码科智能】万物皆可描述！遥感图像描述进入对象级细粒度理解时代|【CV炼丹术】DescribeEarth：西安交大最新论文——面向遥感图像的细粒度对象描述新范式！|