Fire and Smoke Dataset

Dataset Features

• Dataset Size:
  Contains more than 7,000 high-quality images of fire and smoke.
• Contributors:
  Collected by over 1,000 contributors to ensure diversity and coverage.
• Resolution:
  Approximately 98% of the images are in HD resolution (1920×1080 and above).
• Geographical Coverage:
  Images are captured across more than 400 cities in India.
• Diversity:
  Includes images taken in different lighting conditions such as day and night, along with varied angles and distances.
• Devices Used:
  Captured using mobile devices between 2020 and 2021.

Available Annotation Formats

The dataset supports multiple annotation formats for flexibility:

• COCO
• YOLO
• PASCAL VOC
• TFRecord

Conclusion

This dataset provides a strong foundation for developing machine learning models focused on fire and smoke detection in real-world environments. With its diverse and high-quality images, it supports the creation of reliable AI systems for early warning, smart surveillance, and safety monitoring. Additionally, it enables improved model performance across varying conditions, making it a valuable resource for building scalable and effective computer vision solutions.

This dataset is sourced from Kaggle.