SPECIAL: Zero-shot Hyperspectral Image Classification With CLIP

---

Zero-shot hyperspectral classification framework powered by CLIP

---

🛠️ Code Usage Guide

1. Environment Setup

# Create conda environment
conda create -n special python=3.10 
conda activate special

# Install core dependencies
pip install requirements.txt

2. Data Preparation

Download the data from Baidu Netdisk(code: y15t) and place it in the data folder. (The Reflectance of AeroRIT needs to be downloaded from GoogleDrive)

3. Download CLIP

Download the CLIP from Baidu Netdisk(code: vii1) and place it in the checkpoint folder.

4. Run Classification

# AeroRIT
python test.py --data_name AeroRIT --epoches 10 --warmup_epoches 5 --eta_min 1e-5

# Pavia Centre
python test.py --data_name PaviaC --epoches 20 --warmup_epoches 10 --eta_min 1e-4

🙌 Acknowledgements

This work builds upon FeatUp and SegEarth-OV, and we extend our sincere gratitude to the researchers and developers.

📧 Email

[email protected]