基于半监督学习的多场景火灾小规模稀薄烟雾检测

杨凯博,钟铭恩,谭佳威,邓智颖,周梦丽,肖子佶

Small-scale sparse smoke detection in multiple fire scenarios based on semi-supervised learning

Kaibo YANG,Mingen ZHONG,Jiawei TAN,Zhiying DENG,Mengli ZHOU,Ziji XIAO

表 7 不同烟雾集下各个模型的性能对比

Tab.7 Performance comparison of each model with different smoke sets

算法 Easy Hard_1 Hard_2 MSIFSD D-Fire FLOPs/109 FPS AP/% R/% AP/% R/% AP/% R/% AP/% R/% AP/% R/% Faster R-CNN[30] 83.6 73.2 73.6 65.7 77.1 68.7 79.1 69.3 73.2 62.9 40.3 25.3 SSD[31] 75.4 62.8 67.3 57.7 68.2 58.4 71.3 59.4 68.5 64.3 27.6 51.1 YOLOv5n[32] 89.5 81.3 78.2 72.0 82.1 75.5 84.9 77.2 80.3 74.2 1.7 57.8 YOLOv8n 95.0 88.4 82.3 77.3 83.4 78.3 89.9 83.7 86.2 81.4 8.2 103.0 DETR[33] 75.6 85.4 69.9 77.6 71.3 79.2 72.2 81.6 69.3 76.6 8.7 52.1 DeepSmoke 96.3 89.6 86.4 81.0 87.8 82.3 92.0 85.6 88.3 84.1 9.3 85.3 DeepSmoke_SST 98.3 91.4 88.2 82.7 90.0 84.3 94.2 87.6 92.6 88.3 9.3 85.3