Despite the large surges in technology investment and the dominance of smart systems, some emergency events remain unsolved or solved with a massive trade-off between performance and portability. The sudden fire outbreak that may occur due to several reasons is one of the problems that is yet to be solved, a natural disaster that could cause massive ecological and economic losses. Therefore, it is crucial to develop precise, fast, and commercially available solutions. Several image classification approaches were effectively designed, However, most of these models contain many layers that often lead to a massive model size such as AlexNet and GoogleNet. In this paper, we propose a model inspired by FireNet that balances and softens the tradeoff between performance and model size. In this work, smoke detection is augmented to FireNet, the model is trained and tested on various datasets and has an accuracy of 83% which is quite promising as compared to other CNN models that are deployed on an embedded device such as raspberry pi.
%0 Conference Paper
%1 9694193
%A Shoukry, Nadeen
%A Ehab, Farah
%A Salem, Mohammed A.-M
%B 2021 Tenth International Conference on Intelligent Computing and Information Systems (ICICIS)
%D 2021
%K detection myown
%P 66-73
%R 10.1109/ICICIS52592.2021.9694193
%T An Improved Deep Learning Model for Early Fire and Smoke Detection on Edge Vision Unit
%X Despite the large surges in technology investment and the dominance of smart systems, some emergency events remain unsolved or solved with a massive trade-off between performance and portability. The sudden fire outbreak that may occur due to several reasons is one of the problems that is yet to be solved, a natural disaster that could cause massive ecological and economic losses. Therefore, it is crucial to develop precise, fast, and commercially available solutions. Several image classification approaches were effectively designed, However, most of these models contain many layers that often lead to a massive model size such as AlexNet and GoogleNet. In this paper, we propose a model inspired by FireNet that balances and softens the tradeoff between performance and model size. In this work, smoke detection is augmented to FireNet, the model is trained and tested on various datasets and has an accuracy of 83% which is quite promising as compared to other CNN models that are deployed on an embedded device such as raspberry pi.
@inproceedings{9694193,
abstract = {Despite the large surges in technology investment and the dominance of smart systems, some emergency events remain unsolved or solved with a massive trade-off between performance and portability. The sudden fire outbreak that may occur due to several reasons is one of the problems that is yet to be solved, a natural disaster that could cause massive ecological and economic losses. Therefore, it is crucial to develop precise, fast, and commercially available solutions. Several image classification approaches were effectively designed, However, most of these models contain many layers that often lead to a massive model size such as AlexNet and GoogleNet. In this paper, we propose a model inspired by FireNet that balances and softens the tradeoff between performance and model size. In this work, smoke detection is augmented to FireNet, the model is trained and tested on various datasets and has an accuracy of 83% which is quite promising as compared to other CNN models that are deployed on an embedded device such as raspberry pi.},
added-at = {2022-10-11T14:42:34.000+0200},
author = {Shoukry, Nadeen and Ehab, Farah and Salem, Mohammed A.-M},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/26d79f42dc500c217147aad095d175b6d/nfathallah},
booktitle = {2021 Tenth International Conference on Intelligent Computing and Information Systems (ICICIS)},
doi = {10.1109/ICICIS52592.2021.9694193},
interhash = {cefcb8bc564d83efdeb473dac1cc7fcc},
intrahash = {6d79f42dc500c217147aad095d175b6d},
keywords = {detection myown},
month = dec,
pages = {66-73},
timestamp = {2022-10-24T14:49:22.000+0200},
title = {An Improved Deep Learning Model for Early Fire and Smoke Detection on Edge Vision Unit},
year = 2021
}