Satellite Disaster Forecast Prototype
Budget: ₹600 – ₹1,500 INR
I need a compact proof-of-concept that shows how deep learning can predict and forecast floods, wildfires, and earthquakes by combining satellite imagery with timely insights mined from public social-media feeds.
For this initial prototype, keep the scope lean:
• Collect a small, representative sample of open satellite images and matching social posts.
• Build a lightweight preprocessing pipeline to align the two data streams in near–real time.
• Train a basic model—CNN, CNN-LSTM, or another suitable architecture—that outputs short-term risk levels for each disaster type.
• Package the work in a single Jupyter notebook (or equivalent script) with clear comments so I can rerun it locally.
• Include a brief README that explains data sources, key steps, and how to extend the approach later.
Accuracy can be modest at this stage; the goal is to demonstrate feasibility and lay the groundwork for a fuller system once I secure a higher budget.
For this initial prototype, keep the scope lean:
• Collect a small, representative sample of open satellite images and matching social posts.
• Build a lightweight preprocessing pipeline to align the two data streams in near–real time.
• Train a basic model—CNN, CNN-LSTM, or another suitable architecture—that outputs short-term risk levels for each disaster type.
• Package the work in a single Jupyter notebook (or equivalent script) with clear comments so I can rerun it locally.
• Include a brief README that explains data sources, key steps, and how to extend the approach later.
Accuracy can be modest at this stage; the goal is to demonstrate feasibility and lay the groundwork for a fuller system once I secure a higher budget.