Data Sets

View the evaluation dataset here

Evaluation dataset

Knowledge Base Dataset

Overview

Our system leverages a multimodal approach by converting diverse environmental data—including satellite imagery, UAV thermal scans, weather sensor readings, and audio recordings—into high-dimensional vector embeddings .These vectorized representations are stored in a vector databases, enabling efficient semantic search and retrieval of relevant patterns for wildfire detection. When a query is received, our Multi-Agent System retrieves the most contextually similar vectors from the database, which are then processed by Large Content Models to generate predictions about fire risks, spread patterns, or detection alerts while maintaining explainability through the ontology-based knowledge graph.

Algerian Forest Fires

Link : https://archive.ics.uci.edu/dataset/547/algerian+forest+fires+dataset

The dataset includes 244 instances that regroup data of two regions of Algeria, namely the Bejaia region, located in the northeast of Algeria, and the Sidi Bel-Abbes region, located in the northwest of Algeria. 122 instances for each region. The period from June 2012 to September 2012. The dataset includes 11 attributes and 1 output attribute.

Remote Sensing Data Before and After California Rim and King Forest Fires, 2010-2015

Link : https://www.earthdata.nasa.gov/data/catalog/ornl-cloud-king-rim-fire-analysis-1288-1

This data set provides high-resolution surface reflectance, thermal imagery, burn severity metrics, and LiDAR-derived structural measures of forested areas in the Sierra Nevada Mountains, California, USA, collected before and after the August 2013 Rim and September 2014 King mega forest fires. Pre-fire data were paired with post-fire collections to assess pre- and post-fire landscape characteristics and fire severity. Field estimates of fire severity were collected to compare with derived remote sensing indices. Reflectance measurements for the spectroscopic AVIRIS and MASTER sensors are distributed as multi-band geotiffs for each megafire and acquisition date. Derived operational metric products for each sensor are provided in individual GeoTIFFs.

Forest fires, in the northeast region of Portugal

link : https://archive.ics.uci.edu/dataset/162/forest+fires

The dataset includes 517 instances of data in the northeast region of Portugal. Contains tabular data for date, FFMC, DMC, DC, ISI, temperature, and relative humidity.

Wildfire satellite images in Canada

Link : https://www.kaggle.com/datasets/abdelghaniaaba/wildfire-prediction-dataset

Satellite images of areas that previously experienced wildfires in Canada. This contains the images with wildfire and non-wildfire images.

Boreal Forest Fire

Link : https://etsin.fairdata.fi/dataset/1dce1023-493a-4d63-a906-f2a44f831898/data

UAV images & This dataset consists of annotated images of smoke resulting from prescribed burning events in Finnish boreal forests. The data was captured at four events in Evo, Heinola, Karkkila, and Ruokolahti, using a DJI Phantom 4 drone.

FLAME 3 - Radiometric Thermal UAV Imagery for Wildfire Management

Link : https://ieee-dataport.org/open-access/flame-3-radiometric-thermal-uav-imagery-wildfire-management

Included are 622 image quartets labeled Fire and 116 image quartets labeled No Fire. The No Fire images are of the surrounding forestry of the prescribed burn plot. Each image quartet is composed of four images - a raw RGB image, a raw thermal image, a corrected FOV RGB image, and a thermal TIFF. This dataset also contains a NADIR Thermal Fire set.

Forest Wild Fire Sound Dataset

Link : https://www.kaggle.com/datasets/forestprotection/forest-wild-fire-sound-dataset

This dataset contains around 280 audio samples, each of duration 50 seconds.

Page updated

Report abuse