Wildfire Prediction Dataset
This dataset contains satellite images extracted around areas of forest fires detected in Canada. The images are classified into two categories: fire or absence of fire. Ideal for training AI fire prediction and detection models.
Approximately 42,850 satellite images (350x350 px), image format
Creative Commons 4.0 Attribution (CC-BY)
Description
The “Wildfire Prediction Dataset” dataset offers more than 42,000 satellite images (350x350 pixels) divided into two classes: fire zones and fire free zones. These images were extracted from GPS coordinates of identified fires in Canada, using the MapBox API.
What is this dataset for?
- Training computer vision models for forest fire detection and prediction
- Develop environmental monitoring systems based on satellite images
- Test multispectral image classification and spatial analysis algorithms
Can it be enriched or improved?
The dataset can be enriched with additional temporal data to track the progress of fires, or by more accurate annotations on fire intensity and surrounding vegetation. Integration with other multispectral satellite sources could improve the robustness of the models.
🔎 In summary
🧠 Recommended for
- Environmental researchers
- AI developers in satellite vision
- Geospatial image analysis projects
🔧 Compatible tools
- TensorFlow
- PyTorch
- Keras
- OpenCV
- Google Earth Engine
💡 Tip
Combine with weather data to improve predictive accuracy.
Frequently Asked Questions
Is this dataset suitable for the real-time detection of fires?
This dataset is mainly used for model training, but it can be integrated into detection systems in near real time with up-to-date satellite images.
Can I use this dataset for a commercial project?
Yes, the CC-BY 4.0 license allows commercial use, provided the source is mentioned.
Are the images multispectral or only RGB?
The dataset is composed of standard RGB satellite images, with no additional multispectral data.
