Fire & Smoke Detection Dataset (40K Images)

A large-scale, annotated object detection dataset containing over 40,900 images dedicated to early fire and smoke detection. Designed for training real-time vision models such as YOLO (Ultralytics), RT-DETR, and Vision Transformers.

---

Dataset Summary

• Total Images: ~40,900 images
• Task: Object Detection (object-detection)
• Bounding Box Format: YOLO format (class_id x_center y_center width height) / COCO format
• Target Classes:
  • 0: Fire — Active flames and embers
  • 1: Smoke — Smoke plumes and rising smoke

---

Dataset Structure & Splits

The dataset is partitioned into standard training, validation, and testing subsets:

---

Directory Structure (YOLO Format)

dataset/
├── data.yaml
├── train/
│   ├── images/
│   └── labels/
├── valid/
│   ├── images/
│   └── labels/
└── test/
    ├── images/
    └── labels/

Example data.yaml Configuration

​To train Ultralytics YOLO models (e.g., YOLOv8, YOLOv9, YOLOv11) directly with this dataset:

path: ./dataset # Dataset root directory
train: train/images
val: valid/images
test: test/images

Classes names:

0: Fire 1: Smoke

How to use:

With Ultralytics:

pip install ultralytics

from ultralytics import YOLO

# Load a pretrained base model
model = YOLO("yolov8s.pt")

# Train the model
results = model.train(
    data="path/to/data.yaml",
    epochs=100,
    imgsz=800,
    batch=64,
    optimizer="AdamW",
    lr0=0.01
)

With HuggingFace:

from datasets import load_dataset

# Load dataset from Hugging Face
dataset = load_dataset("jojomoi-meme/fog_fire_detection")

# Inspect a sample
print(dataset["train"][0])

Source

• License: MIT
• Compiled and processed from the Roboflow Universe Fire Detection Dataset collections.

Citation & Attribution:

If you use this dataset in a research paper, open-source project, or commercial application, please consider citing it as follows:

@dataset{fire_smoke_detection_dataset_2024,
  author       = {Joachim Servant},
  title        = {Fire and Smoke Detection Dataset},
  year         = {2026},
  publisher    = {Hugging Face Datasets},
  howpublished = {https://huggingface.co/datasets/jojomoi-meme/fog_fire_detection},
  note         = {Dataset containing 40,900 images for real-time fire and smoke detection}
}