
Experimental Exploration of Compact Convolutional Neural Network Architectures for Non-temporal Real-time Fire Detection (G. Samarth, N. Bhowmik, T.P. Breckon), In Proc. Int. Conf. on Machine Learning Applications, IEEE, 2019.

-----------------

@InProceedings{samarth19fire,
  author = 	 {Samarth, G. and Bhowmik, N. and Breckon, T.P.},
  title = 	 {Experimental Exploration of Compact Convolutional Neural Network Architectures for Non-temporal Real-time Fire Detection},
  booktitle =   {Proc. Int. Conf. on Machine Learning Applications},
  year = 	 {2019},
  month = 	 {December},
  publisher =    {IEEE},
  keywords =     {fire detection, CNN, deep-learning real-time, non-temporal},
}
-----------------

﻿Dataset collected from various online sources, with a limited excerpt
of any given source included here-in. Collated and used under
fair use terms for academic research purposes:
(https://www.gov.uk/guidance/exceptions-to-copyright#fair-dealing).

-----------------

The dataset provides png images of non-fire superpixels only. These are taken from online video sources.

-----------------

The data repository structure is as follows:

├── non-fire-superpixels (186Mb / 46,890 PNG format files - the additional training data sets for the superpixel fire localization task) **
└── readme.txt

Each PNG image contains one superpixel, centred in the image.

** used for training /testing on the superpixels fire localization task as reported in the study [Samarth, Bhowmik, Breckon - 2019] as a supplement to the original [Dunnings, Breckon 2018] dataset (available here - http://doi.org/10.15128/r19880vq98m).

For the [Samarth, Bhowmik, Breckon - 2019] study the superpixel training data set was constructed as follows:
- 8,635 fire superpixel images for (from [Dunnings, Breckon 2018] dataset)
- 10,000 non-fire superpixel images (8000 randomly selected from [Dunnings, Breckon 2018] and 2000 randomly selected from this set)

All other datasets (for superpixels + full frame binary detection) for the [Samarth, Bhowmik, Breckon - 2019] paper were from the original [Dunnings, Breckon 2018] dataset.

-----------------

A reference copy of the convolutional neural network models used in this study [Samarth, Bhowmik, Breckon - 2019] is available at: https://github.com/tobybreckon/fire-detection-cnn

If using this dataset for commercial or non-commercial purposes, an acknowledgement must be made in supporting materials to: [Samarth, Bhowmik, Breckon - 2019] in some form.
For academic work - please reference the above paper. For commercial work - provide a reference in the associated readme/documentation/license info that is visible to the end use of the software.

-----------------
