Add model card for Pi-Seg (image-segmentation)

README.md

@@ -1,3 +1,42 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+pipeline_tag: image-segmentation
+---
+
+# Pi-Seg (Perturbation-Injected Segmentation)
+
+This repository contains the weights for **Pi-Seg**, a baseline model for open-vocabulary remote sensing image segmentation (OVRSIS) presented in the paper [Towards Realistic Open-Vocabulary Remote Sensing Segmentation: Benchmark and Baseline](https://huggingface.co/papers/2604.15652).
+
+## Overview
+
+Pi-Seg introduces a **positive-incentive noise** learning mechanism built on the CAT-Seg framework. It learns a broader, more transferable feature space for remote sensing images by injecting semantically guided perturbations into both visual and textual feature spaces during training. This approach significantly improves the model's generalization to unseen categories and novel domains without requiring heavy multi-encoder frameworks.
+
+- **Paper**: [Towards Realistic Open-Vocabulary Remote Sensing Segmentation: Benchmark and Baseline](https://huggingface.co/papers/2604.15652)
+- **Repository**: [Official Pi-Seg GitHub Codebase](https://github.com/LiBingyu01/Pi-Seg)
+- **Datasets**: 
+  - [OVRSIS95K Training Dataset](https://huggingface.co/datasets/kkk2026/OVRSIS95K)
+  - [OVRSISBenchV2 Test Dataset](https://huggingface.co/datasets/kkk2026/OVRSISBenchtest)
+  - [OVRSISBenchV2 Other Tasks](https://huggingface.co/datasets/kkk2026/OVRSISBenchV2o3)
+
+## Citation
+
+If you find this work useful, please cite:
+
+```bibtex
+@article{li2026towards,
+  title={Towards Realistic Open-Vocabulary Remote Sensing Segmentation: Benchmark and Baseline},
+  author={Li, Bingyu and Huo, Tao and Dong, Haocheng and Zhang, Da and Zhao, Zhiyuan and Gao, Junyu and Li, Xuelong},
+  journal={arXiv preprint arXiv:2604.15652},
+  year={2026}
+}
+
+@inproceedings{li2026exploring,
+  title={Exploring efficient open-vocabulary segmentation in the remote sensing},
+  author={Li, Bingyu and Dong, Haocheng and Zhang, Da and Zhao, Zhiyuan and Sun, Hao and Gao, Junyu},
+  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
+  volume={40},
+  number={8},
+  pages={5982--5991},
+  year={2026}
+}
+```