A fire monitoring and alarm system based on channel-wise pruned YOLOv3

Qian, Huimin; Shi, Fei; Chen, Wei; Ma, Yilong; Huang, Min

doi:10.1007/s11042-021-11224-0

Cited by 11 publications

(5 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Shen et al [ 26 ] used the YOLOv1 model to achieve flame detection, but there is still much room for improvement. Qian [ 27 ] et al introduced channel-wise pruning technology to reduce the number of parameters in YOLOv3, making it more suitable for fire monitoring systems. Wang et al [ 28 ] proposed a lightweight detector, Light-YOLOv4, which considers the balance between performance and efficiency and has good detection performance and speed in embedded scenarios.…”

Section: Related Workmentioning

confidence: 99%

Fire Detection in Ship Engine Rooms Based on Deep Learning

Zhu

Zhang

Wang

et al. 2023

Sensors

View full text Add to dashboard Cite

Ship fires are one of the main factors that endanger the safety of ships; because the ship is far away from land, the fire can be difficult to extinguish and could often cause huge losses. The engine room has many pieces of equipment and is the principal place of fire; however, due to its complex internal environment, it can bring many difficulties to the task of fire detection. The traditional detection methods have their own limitations, but fire detection using deep learning technology has the characteristics of high detection speed and accuracy. In this paper, we improve the YOLOv7-tiny model to enhance its detection performance. Firstly, partial convolution (PConv) and coordinate attention (CA) mechanisms are introduced into the model to improve its detection speed and feature extraction ability. Then, SIoU is used as a loss function to accelerate the model’s convergence and improve accuracy. Finally, the experimental results on the dataset of the ship engine room fire made by us shows that the mAP@0.5 of the improved model is increased by 2.6%, and the speed is increased by 10 fps, which can meet the needs of engine room fire detection.

show abstract

Section: Related Workmentioning

confidence: 99%

Fire Detection in Ship Engine Rooms Based on Deep Learning

Zhu

Zhang

Wang

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Qian et al. [18] performs channel pruning on YOLOv3, and incorporates the method of handling sample imbalance into the training process, thereby designing a fire monitoring and alarm system. The model further improves the detection accuracy while reducing the amount of parameters.…”

Section: Related Workmentioning

confidence: 99%

“…Ghosh et al [17] uses multiple Region Proposal Networks of different sizes to generate RoIs in the Faster R-CNN model, which can detect vehicles of different sizes under different weather conditions. Qian et al [18] performs channel pruning on YOLOv3, and incorporates the method of handling sample imbalance into the training process, thereby designing a fire monitoring and alarm system. The model further improves the detection accuracy while reducing the amount of parameters.…”

Section: Related Workmentioning

confidence: 99%

MPGI‐Terminal defect detection based on M‐FRCNN

Liu

Wang

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

Micro‐precision glass insulated terminal (MPGI‐Terminal) is a high‐precision and widely used micro‐precision component. The MPGI‐Terminal is small in size with 2–5 mm in diameter. It is difficult to distinguish missing blocks on the surface during manual inspection, which is prone to false detection and missing detection. The advantages of deep learning methods in the field of object detection have gradually highlighted. They can be used to construct the missing block detection model of MPGI‐Terminal. A Max‐Min Intersection over Union (MIoU) loss function based on overlap area, distance and aspect ratio is proposed, which helps the model to perform bounding box regression more reasonably, so as to obtain the object position more accurately. In addition, the Spatial Attention Module (SAM) and the Region of Interest Alignment (RoI Align) module are introduced to further improve the model performance. Based on the above modules, a MPGI‐Terminal missing block detection model based on M‐FRCNN is constructed. The results show that the proposed model can meet the requirements of fast and high accuracy of MPGI‐Terminal missing block detection.

show abstract

“…There are two types of imagebased fire object instance detection algorithms: traditional and new generation. Traditional algorithms include R-CNN [37], R-FCN [37], SSD [37], Faster-RCNN [38], YOLOV3 [39,40], YOLOV4 [41][42][43], and YOLOV5 [44]. The new-generation algorithms are as follows.…”

Section: Introductionmentioning

confidence: 99%

Image-Based Fire Detection Using Dynamic Threshold Grayscale Segmentation and Residual Network Transfer Learning

Li,

Sun

2023

Mathematics

View full text Add to dashboard Cite

Research on fire detection based on deep learning has been increasing lately, and current models differ in accuracy and computational complexity. To reduce the loss of semantic information and improve prediction accuracy, this paper combines dynamic threshold grayscale segmentation and residual network transfer learning. An effective training data extension strategy integrates the experimental data of a real fire platform and a forest-image-based fire database, where the experimental data of the real fire platform are shot based on different lighting and backgrounds. It has been proven that this is a good solution that can effectively solve the problem of training robust deep models relying on datasets with large diversity, which can improve the overall performance of the training model. Then, we introduce a network architecture based on dynamic threshold grayscale segmentation and residual net transfer learning, which reaches residual net level accuracy on image-based fire detection tasks with less computation. We train the model with a self-built image dataset, test it on a general fire image dataset, and evaluate it on a self-built fire video dataset. In numerous experiments, our method produced a training and testing accuracy of 98.26% and 99.90%, respectively, with corresponding losses of 0.036 and 0.04. The evaluation accuracy was 90% under the condition of a complex sample environment. Compared with other models, our method has advantages in accuracy, training loss, and cost time. Comprehensive comparisons show effectiveness and improved overall performance.

show abstract

A fire monitoring and alarm system based on channel-wise pruned YOLOv3

Cited by 11 publications

References 34 publications

Fire Detection in Ship Engine Rooms Based on Deep Learning

Fire Detection in Ship Engine Rooms Based on Deep Learning

MPGI‐Terminal defect detection based on M‐FRCNN

Image-Based Fire Detection Using Dynamic Threshold Grayscale Segmentation and Residual Network Transfer Learning

Contact Info

Product

Resources

About