Inspiration
Fires can cause significant damage, and rapid response is essential for saving lives and minimizing property loss. Inspired by the need for a data-driven approach to emergency response, we developed an AI system that estimates fire severity probability based on key factors like distance to the nearest fire station and the time of day.
By leveraging machine learning, we aim to assist emergency responders in prioritizing incidents efficiently, ensuring that resources are allocated optimally.
What It Does
The Fire Map uses AI to predict the probability of fire severity based on: • Distance between the fire and the nearest fire station • Time of day (night vs. day impacts response time and fire spread risk)
The AI model calculates a probability score that estimates how severe a fire might become, helping emergency responders decide how urgently they should respond.
How We Built It
1. Data Collection & Processing:
• We created a dataset using real-world fire station locations and simulated fire incidents.
• Key features include fire station proximity and time of day.
2. Machine Learning Model:
• We used a logistic regression model to estimate the probability of high fire severity.
• The model takes two inputs: distance to the fire station and time of day (day/night).
• The output is a probability score representing the likelihood of severe fire impact.
• The model was trained using scikit-learn and saved with joblib.
3. Backend (Django & Python):
• Django serves API endpoints that accept fire location and time of day.
• The backend loads the trained AI model and returns severity probability scores.
4. Frontend (Folium & Visualization):
• We used Folium to map the fire incident and fire stations dynamically.
• The system visualizes fire severity using color-coded markers (low to high severity).
5. Deployment & Testing:
• The project runs on a local Django server.
• We tested different fire locations and times of day to analyze model predictions.
Challenges We Ran Into
• Simplified Model Scope: Initially, we planned to use more variables like weather and terrain but focused on core factors (distance & time) due to time constraints.
• Geospatial Calculations: Mapping distances dynamically required GIS processing and optimization.
• Model Calibration: Fine-tuning probability thresholds for meaningful severity predictions.
• Deployment Issues: Managing dependencies and ensuring the model runs correctly in the Django backend.
Accomplishments That We’re Proud Of
• Developed a functional AI model that estimates fire severity probability using real-world geospatial data.
• Created an interactive visualization that dynamically maps fire incidents and response units.
• Built a Django-based API that integrates AI predictions into a web-based emergency response tool.
• Successfully handled data processing and distance calculations in real time.
What We Learned
• How to apply logistic regression for probability-based predictions.
• The importance of feature selection (choosing only the most relevant factors).
• How to use Folium for dynamic geospatial visualization.
• How to integrate AI models into Django web applications.
• The challenges of real-world emergency response modeling.
What’s Next for The Fire Map
• Incorporate more real-world factors like weather conditions and fire history.
• Expand the AI model to use more sophisticated machine learning algorithms.
• Deploy the system on the cloud (AWS, GCP, or Heroku) for wider accessibility.
• Integrate live fire incident reporting for real-time emergency response.
• Collaborate with fire departments to enhance practical usability.
Log in or sign up for Devpost to join the conversation.