FIRES is a machine learning–powered web application designed to predict forest fires from satellite imagery. By analyzing environmental and spatial data, FIRES identifies areas at high risk for wildfires, enabling early prevention and saving lives, ecosystems, and infrastructure.
FIRES allows users to upload satellite images of a specific region to determine whether that area is at risk of a wildfire.
The system uses a custom-trained convolutional neural network (MobileNetV2 + MLP) that achieves 97% accuracy on validation data, significantly outperforming traditional risk prediction models.
After processing the input image, FIRES provides:
- A prediction (yes/no) on wildfire risk.
- Actionable safety tips on mitigating or preventing wildfires.
- An analytics dashboard to review past predictions and outcomes.
- Satellite Image Analysis: Upload an image of a region to detect potential wildfire risks.
- High-Accuracy Predictions: 97% detection accuracy with MobileNetV2 + Multi-Layer Perceptron architecture.
- Smart Prevention Tips: Offers tailored recommendations to mitigate wildfire effects.
- Analytics Section: Review historical inputs, outputs, and detection trends.
- Web-Based Interface: Simple, dark-themed, and responsive design for accessibility across devices.
| Category | Technologies |
|---|---|
| Frontend | CSS / Python (Streamlit) |
| Backend | Python |
| Machine Learning | TensorFlow, Keras, NumPy |
| Model Architecture | MobileNetV2 + MLP Classifier |
| Deployment | Localhost / Streamlit Cloud |
| Metric | Result |
|---|---|
| Accuracy | 97% |
| Loss | 0.08 |
| Model Type | Binary Classification (Fire / No Fire) |
-
Upload Satellite Image
The user drags and drops a file for analysis. -
Model Prediction
The trained model processes the image to determine if the area is at risk of a wildfire. -
Results Display
FIRES provides the prediction result and customized tips based on the analysis. -
Analytics Dashboard
Users can view consolidated data from previous predictions to track changes over time.
- Integrate additional data sources (temperature, humidity, vegetation index) to improve accuracy.
- Add LLM integration (e.g., ChatGPT API) for enhanced feedback and interpretability.
- Build a mobile version with real-time push notifications for early warnings.
- Deploy FIRES on a public cloud platform for broader accessibility.
FIRES was inspired by the increasing frequency of wildfires and the devastating effects of climate-related disasters. After personally experiencing severe environmental damage in my own community, I realized the need for an accessible, predictive tool to help prevent tragedies before they happen.
Developer: Varun Nukala
Email: [varun.k.nukala@gmail.com]
Project: Congressional App Challenge Submission 2025
Location: Helotes, Texas