Assessing the Impact of the Loss Function and Encoder Architecture for Fire Aerial Images Segmentation Using Deeplabv3+

被引:20
作者
Harkat, Houda [1 ,2 ]
Nascimento, Jose M. P. [3 ]
Bernardino, Alexandre [4 ]
Ahmed, Hasmath Farhana Thariq [5 ]
机构
[1] Inst Super Tecn, Inst Telecomunicacoes, Av Rovisco Pais 1, P-1049001 Lisbon, Portugal
[2] Univ Sidi Mohamed Ben Abdellah, Fac Sci & Technol, BP 2626,Route Imouzzer, Fes 30000, Morocco
[3] Inst Super Engn Lisboa, IPL, P-1049001 Lisbon, Portugal
[4] ISR Inst Sistemas & Robot, Av Rovisco Pais 1, P-1049001 Lisbon, Portugal
[5] Saveetha Univ, Saveetha Sch Engn, Saveetha Inst Med & Tech Sci, Dept Comp Sci Engn, Chennai 602105, Tamil Nadu, India
关键词
fire; firefront_gestosa; deep learning; deeplabv3+; backbone; dice loss; image processing; NETWORK;
D O I
10.3390/rs14092023
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
Wildfire early detection and prevention had become a priority. Detection using Internet of Things (IoT) sensors, however, is expensive in practical situations. The majority of present wildfire detection research focuses on segmentation and detection. The developed machine learning models deploy appropriate image processing techniques to enhance the detection outputs. As a result, the time necessary for data processing is drastically reduced, as the time required rises exponentially with the size of the captured pictures. In a real-time fire emergency, it is critical to notice the fire pixels and warn the firemen as soon as possible to handle the problem more quickly. The present study addresses the challenge mentioned above by implementing an on-site detection system that detects fire pixels in real-time in the given scenario. The proposed approach is accomplished using Deeplabv3+, a deep learning architecture that is an enhanced version of an existing model. However, present work fine-tuned the Deeplabv3 model through various experimental trials that have resulted in improved performance. Two public aerial datasets, the Corsican dataset and FLAME, and one private dataset, Firefront Gestosa, were used for experimental trials in this work with different backbones. To conclude, the selected model trained with ResNet-50 and Dice loss attains a global accuracy of 98.70%, a mean accuracy of 89.54%, a mean IoU 86.38%, a weighted IoU of 97.51%, and a mean BF score of 93.86%.
引用
收藏
页数:22
相关论文
共 50 条
[1]  
Akhloufi M.A., P PATT REC TRACK 29
[2]  
[Anonymous], 2017, J SCI ENG RES
[3]  
Badiger V., 2020, SAMRIDDHI J PHYS SCI, V12, P50
[4]   Low-Power Distributed Sensor Network for Wildfire Detection [J].
Blalack, Triston ;
Ellis, Dakota ;
Long, Marcus ;
Brown, Carter ;
Kemp, Ricardo ;
Khan, Muhammad .
2019 IEEE SOUTHEASTCON, 2019,
[5]   Wireless Sensor Network for Ignitions Detection: An IoT approach [J].
Brito, Thadeu ;
Pereira, Ana, I ;
Lima, Jose ;
Valente, Antonio .
ELECTRONICS, 2020, 9 (06)
[6]   Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation [J].
Chen, Liang-Chieh ;
Zhu, Yukun ;
Papandreou, George ;
Schroff, Florian ;
Adam, Hartwig .
COMPUTER VISION - ECCV 2018, PT VII, 2018, 11211 :833-851
[7]   Smoke detection and trend prediction method based on Deeplabv3+and generative adversarial network [J].
Cheng, Shuhong ;
Ma, Jiyong ;
Zhang, Shijun .
JOURNAL OF ELECTRONIC IMAGING, 2019, 28 (03)
[8]   Semantic Fire Segmentation Model Based on Convolutional Neural Network for Outdoor Image [J].
Choi, Han-Soo ;
Jeon, Myeongho ;
Song, Kyungmin ;
Kang, Myungjoo .
FIRE TECHNOLOGY, 2021, 57 (06) :3005-3019
[9]  
Chollet F., 2017, P IEEE C COMPUTER VI, V1251, P1258
[10]   Sensitivity of Portuguese forest fires to climatic, human, and landscape variables: subnational differences between fire drivers in extreme fire years and decadal averages [J].
Costa, Luis ;
Thonicke, Kirsten ;
Poulter, Benjamin ;
Badeck, Franz-W. .
REGIONAL ENVIRONMENTAL CHANGE, 2011, 11 (03) :543-551