Multimodal Fusion Wildfire & Vision-LLM Suite

A collection of Jupyter notebooks and a Streamlit demo app demonstrating:

1. Late Fusion of RGB & Thermal imagery
   – Late fusion_1.ipynb
   – Combines independent CNN or transformer branches at decision level.

2. Early Fusion Multimodal Network
   – Multimodal Fusion Network_early fusion_1.ipynb
   – Stacks four-channel inputs (RGB+Thermal) and trains a CrossViT backbone.

3. Vision LLM Integration
   – Vision LLM/
   – Code for wrapping model outputs in natural-language explanations via KOSMOS-2 or similar.

---