ACADSTAFF UGM
| Title | : | A Comparison of Deep Learning Methods for Vision-based Fire Detection in Surveillance System |
| Author | : |
Wahyono, Ph.D. (1) Prof. Drs. Agus Harjoko, M.Sc., Ph.D. (2) Dr. Andi Dharmawan, S.Si., M.Cs. (3) Gamma Kosala (4) PUTRA YUDHA PRANATA (5) |
| Date | : | 0 2021 |
| Keyword | : | Deep Learning,Faster RCNN,Fire detection,Yolov4,Yolov5 Deep Learning,Faster RCNN,Fire detection,Yolov4,Yolov5 |
| Abstract | : | Fire detection by analyzing video and images from surveillance cameras is increasingly being used as an early warning system. Fire detection using video surveillance as input is much more efficient than fire sensors because of the broader coverage area and is relatively cheaper. Several previous studies used handcrafted features as a differentiator between fire and other objects. Features like color, texture, and motion can be used as a reference for detecting fire. Object detection techniques using video as input are overgrowing. Some researchers use deep learning architecture to detect objects from video input. The deep learning architecture is capable of extracting features directly from video input by leveraging neural network layers. This study tries to compare several deep learning methods for fire detection from video input. The methods compared are Faster RCNN, Yolov4, and Yolov5. After being evaluated, the method with the best True Positive Rate is Yolov4 with a value of 84.62%, while the method with the best True Negative Rate value is Yolov5 with a value of 97.06%. Yolov5 is also the method with the best computational time that is 23.26 FPS. Faster RCNN got an excellent positive rate, almost close to the positive rate of Yolov4 and Yolov5. However, the negative rate of Faster RCNN is quite alarming. The computation time required is also longer than Yolov4 and Yolov5. |
| Group of Knowledge | : | |
| Level | : | Internasional |
| Status | : |
Published
|
| No | Title | Action |
|---|