Real-time Detection of Flame and Smoke Using an Improved YOLOv4 Network

Wang, Yifan,Hua, Changchun,Ding, Weili,Wu, Ruinan

DOI: https://doi.org/10.1007/s11760-021-02060-8

2022-01-01

Abstract:Fire is one of the major disasters in the world, which seriously endangers the safety of life and property. Effective flame and smoke detection can provide timely warning information for firefighters. Existing flame and smoke detection algorithms are limited by processor performance and cannot operate large deep networks. This paper proposes a lightweight detector called Light-YOLOv4, which considers the balance between performance and efficiency. First, the backbone network CSPDarknet53 is replaced by MobileNetv3. Second, the path aggregation network is changed into bidirectional feature pyramid network (BiFPN) based on the idea of bidirectional cross-scale connections. Third, an efficient feature extraction module named depthwise separable attention module based on depthwise separable convolution and coordinate attention network is construed to replace the 3 $$\times $$ 3 standard convolution of spatial pyramid pooling, BiFPN and YOLO head network. In comparison with YOLOv4, Light-YOLOv4 has only 19.1% of its trainable parameters while almost keeping the same accuracy. By combining multiple tricks, Light-YOLOv4 can achieve a better balance between performance (85.64% mAP) and efficiency (71 FPS), which meets flame and smoke detection tasks’ requirements on the accuracy and real time. Experiments on Nvidia Jetson TX2 further demonstrate that Light-YOLOv4 has good detection performance and speed on embedded scenarios.