합성곱 신경망 기반 화재 인식 모델 최적화 연구: Layer Importance Evaluation 기반 접근법
A Study on the Optimization of Fire Awareness Model Based on Convolutional Neural Network: Layer Importance Evaluation-Based Approach
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 연구는 Layer Importance Evaluation을 통해 도출된 화재 감지에 최적화된 딥러닝 아키텍처를 제안한다. 기존의 합성곱 신경망(ConvolutionalNeural Network, CNN) 기반 화재 감지 시스템의 불필요한 복잡성과 연산을 초래하는 문제점을 해결하기 위해, Layer Importance Evaluation기법을 통해 가중치 및 활성화 값에 근거한 모델의 내부 레이어의 동작을 분석하고, 화재 감지에 기여도가 높은 레이어를 식별한 뒤, 식별한 레이어만으로 모델을 재구성하여, 기존 모델과의 성능 지표를 비교 분석하였다. Xception, VGG19, ResNet, EfficientNetB5 등 네 가지 전이 학습 모델을사용하여 화재 데이터를 학습시킨 후, Layer Importance Evaluation기법을 적용하여 각 레이어의 가중치와 활성화 값을 분석한 뒤 기여도가 가장높은 상위 랭크 레이어들을 선별하여 새로운 모델을 구축하였다. 연구 결과, 구현된 아키텍처는 기존 모델 대비 약 80% 가량 경량화 된 파라미터로도동등한 성능을 유지하며, 약 3~5배가량 신속한 학습 속도를 가지면서도 기존의 복잡한 전이학습 모델에 비해 정확도, 손실, 혼동행렬 지표에서동등한 성능을 출력함으로써, 화재 감시 장비의 효율성을 높이는 데 기여할 수 있음을 확인하였다.
Abstract
This study proposes a deep learning architecture optimized for fire detection derived through Layer Importance Evaluation. In orderto solve the problem of unnecessary complexity and operation of the existing Convolutional Neural Network (CNN)-based fire detectionsystem, the operation of the inner layer of the model based on the weight and activation values was analyzed through the Layer ImportanceEvaluation technique, the layer with a high contribution to fire detection was identified, and the model was reconstructed only withthe identified layer, and the performance indicators were compared and analyzed with the existing model. After learning the fire datausing four transfer learning models: Xception, VGG19, ResNet, and EfficientNetB5, the Layer Importance Evaluation technique was appliedto analyze the weight and activation value of each layer, and then a new model was constructed by selecting the top rank layers withthe highest contribution. As a result of the study, it was confirmed that the implemented architecture maintains the same performancewith parameters that are about 80% lighter than the existing model, and can contribute to increasing the efficiency of fire monitoringequipment by outputting the same performance in accuracy, loss, and confusion matrix indicators compared to conventional complextransfer learning models while having a learning speed of about 3 to 5 times faster.
참고문헌 (16)
- W. Zeng, "Early fire detection using convolutional neural networks for CCTV surveillance cameras", Proceedings of the International Conference on Computer Vision and Pattern Recognition (CVPR), pp.629~638 (2022), IEEE
- S. Han, "Fire detection in video surveillance using convolutional neural networks", International Journal of Advanced Robotic Systems, 18(3), pp.1~12 (2021)
- J. L. Hennessy, "Computer Architecture: A Quantitative Approach", pp.66~153 (2007), Morgan Kaufmann Pub
- "Optimizing Machine Learning Models for Real-Time Data Processing" (2024)[URL]
- Y. LeCun, "Backpropagation applied to handwritten zip code recognition", Neural Computation, 1(4), pp.541~551 (1989)
- Y. Shi, "A comprehensive review of artificial intelligence applications in power systems", Neural Networks, 172, pp.372~393 (2024)
- Y. LeCun, "Optimal brain damage", Advances in Neural Information Processing Systems, pp.598~605 (1990)
- M. Abadi, "TensorFlow: A system for large-scale machine learning", Proceedings of the 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI), pp.265~283 (2016)
- "Fire Dataset" (2020)[URL]
- I. Goodfellow, "Generative adversarial networks", pp.2672~2680 (2014)
- T. Salimans, "Improved techniques for training GANs", Advances in Neural Information Processing Systems, pp.2234~2242 (2016)
- F. Chollet, "Xception : Deep learning with depthwise separable convolutions", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.1251~1258 (2017)
- K. Simonyan, "Very deep convolutional networks for large-scale image recognition", Proceedings of the International Conference on Learning Representations(ICLR) (2015)
- K. He, "Deep residual learning for image recognition", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.770~778 (2016)
- M. Tan, "EfficientNet : Rethinking model scaling for convolutional neural networks", Proceedings of the 36th International Conference on Machine Learning (ICML), pp.6105~6114 (2019)
- W. Jin, "FCPNet : An optimized deep learning architecture for fire detection based on explainable AI", Proceeding of the Annual Symposium of KIPS 2024, 31(1), pp.719~720 (2024)