jojomoi-meme/fog_fire_detection

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:

Split	Number of Images	Description
Train	29,656	Main split used for model training
Validation	7,881	Used for hyperparameter tuning and early stopping
Test	3,363	Unseen benchmarks used for final performance evaluation
Total	40,900	Full dataset size

---

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}
}