Simuletic/Long_Distance_Wildfire_Smoke_Detection_Dataset

Wildfire Smoke Detection Dataset (fire-lookout cameras, YOLO + COCO): free 239-image sample

This wildfire smoke detection dataset is a free sample of 239 synthetic images of distant wildfire smoke seen from elevated fire-watch cameras, with YOLO bounding boxes for smoke and wildfire and a text caption per image, for training and testing wildfire and smoke detection models. Most images show smoke only, which is how an early wildfire usually looks from far away. The full 1,000-image dataset adds night fires with glowing flame lines, COCO labels and about 28% look-alike hard negatives (cloud on ridges, fog, mist, dust trails, steam), with a 900 / 100 train-val split. Made by Simuletic, a Swedish company that has built synthetic CCTV datasets for computer vision since 2023.

Get the full dataset · Free sample licence: CC BY-NC 4.0. This free sample is also on Kaggle.

Facts

	Free sample (this repo)	Full dataset
Images	239 PNG (189 smoke only, 50 with flames)	1,000 (900 train / 100 val); 719 fire scenes, 183 with visible flames, 50 at night
Classes	smoke, wildfire	smoke, wildfire
Boxes	255 smoke, 71 wildfire	753 smoke, 190 wildfire; 1 in 6 plumes under 40 px tall
Hard negatives	none	about 28%: cloud on ridges, fog banks, valley mist, dust trails, steam, rain shafts, glare
Resolution	1,024 × 1,024 px	lookout-camera frames; dawn to dusk and night; clear to hazy
Camera view	elevated long-range views over forest, hills and valleys	mountain-top and mast-mounted lookout cameras; forest, grassland, snow
Labels	YOLO .txt + one JSONL of captions	YOLO + COCO, dataset card
Licence	CC BY-NC 4.0	one-time: research $149 / commercial $590 per 1,000 images

What's in the images

Early wildfires seen from far away, the way a fire-watch camera on a tower or ridge sees them. Smoke usually shows long before flames, and flames are often hidden by terrain or trees, so most images show smoke only: a thin grey trail or a small plume rising from a ridge, valley or hillside. In 50 of the 239 images a small flame or orange glow is visible at the base of the smoke. Plumes are often tiny, low-contrast objects in a wide landscape.

Landscapes in the sample include forested valleys, rocky mountain ridges, dry Mediterranean pine forest, olive groves and scrub, limestone cliffs, dense conifer canopy, eucalyptus savanna and rural land with patches of woodland. The images come from 48 scene setups.

Classes and boxes

Class	YOLO id	What it marks	Boxes in sample
smoke	0	smoke plumes, wisps and columns	255
wildfire	1	visible flames	71

Smoke is boxed in 238 of the 239 images and flames in 47.

Label format and files

Long-Distance  Wildfire & Smoke Dataset/      # note: two spaces after "Long-Distance"
  images/                                     # 239 PNG images (var1_early_wildfire_smoke_only_img1.png, ...)
    wildfire_detection_dataset.jsonl          # one text caption per image
  labels/                                     # one YOLO .txt per image, same file names
  data.yaml                                   # class names: 0 smoke, 1 wildfire

Each YOLO label line is class_id x_center y_center width height, normalised to 0–1. This smoke box (var1_early_wildfire_smoke_only_img1.txt) covers about 4% of the image width:

0 0.492574 0.364480 0.039604 0.068069

wildfire_detection_dataset.jsonl has one line per image in chat format: a fixed question, and an answer with the alert status (smoke only, or flame visible), the location and what is seen. Use it to fine-tune vision-language models that write alert messages:

{"image": "var1_early_wildfire_smoke_only_img1.png", "conversations": [{"role": "user", "content": "Analyze this long-range optical sensor feed for early signs of wildfire."}, {"role": "assistant", "content": "Alert Status: Early Wildfire Smoke Only. Location: a deep forested valley under a clear blue sky. Observation: a subtle trail of grey smoke drifting up from a remote rocky ridge. Only smoke visible."}]}

Quick start (Ultralytics YOLO)

Create a data.yaml next to the dataset folder (quote the path, it contains spaces):

path: "Long-Distance  Wildfire & Smoke Dataset"
train: images
val: images   # small sample: use the same images, or split them yourself
names:
  0: smoke
  1: wildfire

Then train:

from ultralytics import YOLO

model = YOLO("yolo11n.pt")
model.train(data="data.yaml", epochs=30, imgsz=640)

Use cases

• Wildfire early-warning camera networks and AI-assisted lookout towers
• Forest and national park monitoring
• Monitoring utility lines and the edge between towns and wildland
• Vision-language models that describe smoke and fire in an alert, using the captions
• Benchmarking detectors on small, distant, low-contrast objects

Full dataset and prices

The full Long-Distance Wildfire & Smoke Detection Dataset has 1,000 images (900 train / 100 val), YOLO + COCO, classes smoke, wildfire, about 28% hard negatives. Every licence is a one-time payment, perpetual, for one organisation; you get a download link within five working days (card or invoice). Research covers students and non-commercial work and is available for 1,000 images only; packs of 5,000 images and more are sold with the commercial licence, which lets you train models and ship them in products.

	Research	Commercial
1,000 images	$149	$590
5,000-image pack	not offered	$2,400
10,000-image pack	not offered	$4,500
20,000-image pack	not offered	$8,500

Packs are the same dataset at larger size, with the same labels and split, under the commercial licence. 50,000+ images, video datasets and enterprise agreements (from $10,000) are quoted. Other scenes, camera angles or classes: custom data or contact@simuletic.com.

How the images were made

The images are fully synthetic (computer-generated) by Simuletic: no real people, places or incidents are depicted, so there are no privacy or consent issues.

Limitations. Synthetic images do not cover every camera, lens, compression artefact or scene you will meet in deployment. Validate on your own footage, and for production combine synthetic and real data. This sample has 239 images from 48 scene setups; the full dataset has 1,500 images across more biomes, times of day and weather.

More synthetic datasets from Simuletic

• CCTV Smoke & Fire Emergency Detection Dataset
• Airport Drone Threat & Safety Dataset
• CCTV Fall & Incident Detection Dataset
• All datasets: simuletic.com/datasets · Hugging Face · Kaggle · GitHub

Licence

This free sample is licensed under CC BY-NC 4.0: you may share and adapt it for non-commercial purposes, as long as you credit Simuletic. The full dataset is sold under a separate one-time research or commercial licence; it is not redistributed.

Citation

@misc{simuletic_wildfire_2026,
  author       = {{Simuletic}},
  title        = {Long-Distance Wildfire & Smoke Detection Dataset},
  year         = {2026},
  publisher    = {Simuletic},
  howpublished = {\url{https://huggingface.co/datasets/Simuletic/Long_Distance_Wildfire_Smoke_Detection_Dataset}},
  note         = {Free sample. Full dataset: \url{https://simuletic.com/datasets/long-distance-wildfire-smoke-detection-dataset}}
}

Questions or feedback: contact@simuletic.com