改进YOLO v4-tiny的火焰实时检测

计算机工程与科学 ›› 2022, Vol. 44 ›› Issue (12): 2196-2205.

改进YOLO v4-tiny的火焰实时检测

王冠博1，赵一帆2，李波1，杨俊东1，丁洪伟1

(1.云南大学信息学院,云南昆明 650504；2.云南民族大学电气信息工程学院，云南昆明 650031)

收稿日期:2021-08-23 修回日期:2021-09-22 接受日期:2022-12-25 出版日期:2022-12-25 发布日期:2023-01-05
基金资助:
国家自然科学基金（61461053,61461054）

Real-time flame detection with improved YOLO v4-tiny

WANG Guan-bo1,ZHAO Yi-fan2,LI Bo1,YANG Jun-dong1,DING Hong-wei1

(1.School of Information Science & Engineering,Yunnan University,Kunming 650504;
2.School of Electrical and Information Engineering,Yunnan University for Nationalities,Kunming 650031,China)

Received:2021-08-23 Revised:2021-09-22 Accepted:2022-12-25 Online:2022-12-25 Published:2023-01-05

摘要/Abstract

摘要： 为解决火焰实时检测参数量大、对硬件计算能力要求高等问题，提出了改进型YOLO v4-tiny的轻量级火焰实时检测模型。首先，对模型的参数进行了修剪；其次，通过在模型的浅层加入改进型的CSP-RFBs，扩大网络浅层的感受野；然后，对CSP-ResNet的框架进行改进，提出了速度更快、准确率更高的沙漏型CSP-ResNet；最后，在网络深层采用改进型CSP-SPPs，对多重感受野进行进一步融合。实验结果表明，改进型YOLO v4-tiny模型的准确率可达48.5%，较原模型提升了15.5%；模型的参数量和权重文件大小分别为2.45 BFLOPs和16.3 Mb，分别比原模型减少了63.9%和30.6%。在移动开发板NVIDIA Jeston Xavier上FPS可达49.6，比原模型提升了21.9%。

关键词: 轻量级目标检测, YOLO v4-tiny, 火焰实时检测, 感受野, 沙漏型CSP-ResNet

Abstract: In order to solve the problems of large number of parameters for real-time flame detection and high requirements for hardware computing power, a lightweight real-time flame detection model based on improved YOLO v4-tiny is proposed. Firstly, the parameters of the model are pruned. Secondly, by adding improved Receptive Field Blocks (CSP-RFBs) in the shallow layer of the model, the perceptual field of the model shallow layer is improved. Thirdly, the framework of CSP-ResNet is improved, and the “hourglass CSP-ResNet” with faster and higher accuracy is proposed. Finally, a modified Spatial Pyramid Pooling (SPP) is adopted at the deep level of the model to further fuse the multiple sensory fields. The experimental results show that the accuracy of the improved YOLO v4-tiny model can reach 48.5%, which is 15.5% better than the original model. The number of parameters of the model and the weight size of the weight file are 2.45BFLOPs and 16.3Mb, which are 63.9% and 30.6% less than the original model, respectively. The FPS on the mobile development board NVIDIA Jeston Xavier can reach 49.6, which is 21.9% better than the original model.

Key words: lightweight object detection, YOLO v4-tiny, real-time flame detection, receptive field, hourglass CSP-ResNet

王冠博, 赵一帆, 李波, 杨俊东, 丁洪伟. 改进YOLO v4-tiny的火焰实时检测[J]. 计算机工程与科学, 2022, 44(12): 2196-2205.

WANG Guan-bo, ZHAO Yi-fan, LI Bo, YANG Jun-dong, DING Hong-wei. Real-time flame detection with improved YOLO v4-tiny[J]. Computer Engineering & Science, 2022, 44(12): 2196-2205.

参考文献［23］

［1］	Alex K,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks ［J］.Communications of the ACM,2017,60(6):84-90.
［2］	He K M, Zhang X, Ren S, et al.Deep residual learning for image recognition ［C］∥Proc of the IEEE Conference on Computer Vision and Pattern Recognition,2016:770-778.
［3］	Liu W, Anguelov D, Erhan D, et al. SSD:Single shot multibox detector ［C］∥Proc of European Conference on Computer Vision,2016:21-37.
［4］	Redmon J, Divvala S, Girshick R,et al.You only look once:Unified,real-time object detection ［C］∥Proc of the IEEE Conference on Computer Vision and Pattern Recognition,2016:779-788.
［5］	Redmon J, Farhadi A. YOLO9000:Better,faster,stronger ［C］∥Proc of the IEEE Conference on Computer Vision and Pattern Recognition,2017:7263-7271.
［6］	Redmon J, Farhadi A. YOLOv3:An incremental improvement ［J］.arXiv:1804.02767,2018.
［7］	Bochkovskiy A, Wang C Y,Liao H Y M. YOLOv4:Optimal speed and accuracy of object detection ［J］.arXiv:2004.10934,2020.
［8］	Lin T Y, Goyal P, Girshick R, et al.Focal loss for dense object detection ［C］∥Proc of the IEEE International Confe- rence on Computer Vision,2017:2980-2988.
［9］	Chollet F. Xception:Deep learning with depthwise separable convolutions ［C］∥Proc of 2017 IEEE Conference on Computer Vision and Pattern Recognition,2017:1800-1807.
［10］	Howard A G, Zhu M, Chen B, et al.MobileNets:Efficient convolutional neural networks for mobile vision applications ［J］.arXiv:1704.04861,2017.
［11］	Sandler M, Howard A, Zhu M, et al.MobileNetV2:Inverted residuals and linear bottlenecks ［C］∥Proc of 2018 IEEE Conference on Computer Vision and Pattern Recognition,2018:4510-4520.
［12］	Howard A, Sandler M, Chu G,et al.Searching for MobileNetV3 ［C］∥Proc of the IEEE International Conference on Computer Vision,2019:1314-1324.
［13］	Zhang X Y, Zhou X Y, Lin M X,et al.ShuffleNet:An extremely efficient convolutional neural network for mobile devices ［C］∥Proc of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition,2018:6848-6856.
［14］	Ma N N,Zhang X Y,Zheng H T,et al.ShuffleNet V2:Practical guidelines for efficient CNN architecture design ［C］∥Proc of European Conference on Computer Vision,2018:116-131.
［15］	Chen L C,Papandreou G,Kokkinos I,et al.DeepLab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs ［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,40(4):834-848.
［16］	Chen L C,Papandreou G,Schroff F,et al.Rethinking atrous convolution for semantic image segmentation ［J］.arXiv:1706.05587,2017.
［17］	Dai J,Li Y,He K,et al.R-FCN:Object detection via region-based fully convolutional networks ［C］∥Proc of the 30th International Conference on Neural Information Processing Systems,2016:379-387.
［18］	Li Z M,Peng C,Yu G,et al.DetNet:A backbone network for object detection ［J］.arXiv:1804.06215,2018.
［19］	Luo W,Li Y,Urtasun R,et al.Understanding the effective receptive field in deep convolutional neural networks ［C］∥Proc of the 30th International Conference on Neural Information Processing Systems,2016:4898-4906.
［20］	Singh P, Verma V K, Rai P,et al.HetConv:Heterogeneous kernel-based convolutions for deep CNNs ［C］∥Proc of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition,2019:4835-4844.
［21］	Li Y H,Chen Y, Wang N, et al.Scale-aware trident networks for object detection ［C］∥Proc of the IEEE International Conference on Computer Vision,2019:6054-6063.
［22］	Wang C Y, Liao H Y M, Wu Y H,et al.CSPNet:A new backbone that can enhance learning capability of cnn ［C］∥Proc of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops,2020:390-391.
［23］	He K M, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015,37(9):1904-1916.

[1]	李校林, 王复港, 张鹏飞, 张琳玉, . 基于多尺度特征提取的YOLOv5s算法优化[J]. 计算机工程与科学, 2023, 45(06): 1054-1062.
[2]	张海燕, 付应娜, 丁桂江, 孟庆岩. 基于无锚框目标检测算法的多样性感受野注意力特征补偿[J]. 计算机工程与科学, 2022, 44(11): 1995-2002.
[3]	徐世杰, 杜煜, 鹿鑫, 吴思凡. 基于ENet的轻量级语义分割算法研究[J]. 计算机工程与科学, 2021, 43(08): 1454-1460.
[4]	杨国亮,王志元,张雨,康乐乐,胡政伟. 基于垂直区域回归网络的自然场景文本检测[J]. 计算机工程与科学, 2018, 40(07): 1256-1263.