DisasterVLP: Perceiving Multidimensional Disaster Damages from Street-View Images Using Visual–Language Models

🗂 Overview

DisasterVLP is a vision-language framework designed to perceive disaster damages from bi-temporal street-view images. It leverages large language models and contrastive learning to assess disaster severity and generate descriptive captions.

📊 Dataset

Citation:
Yang, Yifan (2025). Perceiving Multidimensional Disaster Damages from Street–View Images Using Visual–Language Models. figshare. Dataset.
https://doi.org/10.6084/m9.figshare.28801208.v2

DisasterVLP: Perceiving Multidimensional Disaster Damages from Street-View Images Using Visual–Language Models

🧠 Overview

DisasterVLP is a vision-language framework that leverages pre- and post-disaster street-view images to perceive and classify disaster damage severity. The model integrates large vision-language models like CLIP, BLIP, and GPT to perform visual-language grounding and disaster impact estimation.

---

📊 Dataset

Yang, Yifan (2025). Perceiving Multidimensional Disaster Damages from Street–View Images Using Visual–Language Models. figshare.
DOI: 10.6084/m9.figshare.28801208.v2

---

🧱 Framework Overview

🔍 Overall Structure

---

🔬 Phase 2: Pseudo Post-Disaster Generation

---

📊 Phase 3: Multi-modal Disaster Perception

---

🌍 Study Area

---

🧪 Label Results Example

---

🚀 Getting Started

🔧 Installation

git clone https://github.com/rayford295/DisasterVLP.git
cd DisasterVLP
pip install -r requirements.txt


## 🔧 Installation

```bash
git clone https://github.com/rayford295/DisasterVLP.git
cd DisasterVLP
pip install -r requirements.txt