Lightweight target detection algorithm based on YOLOv4

Liu, Chuan; Wang, Xianchao; Wu, Qilin; Jiang, Jiabao

doi:10.1007/s11554-022-01251-x

Cited by 10 publications

(7 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Shen et al [17] proposed a semi-supervised infrared image target detection algorithm based on CenterNet and OMix enhancement, used CenterNet as the backbone model to detect target in infrared target images according to key points, and achieved good detection performance by training and testing on public datasets. Liu et al [18] set up the lightweight infrared real-time target detection model MCA-YOLO, which improved the real-time performance of target detection by reducing the model parameters and ensured the detection accuracy. Sun et al [19] proposed a simple and fast infrared image salient target detection algorithm, extracted the spectral residuals in the log spectrum by enhancing the contrast between the target and the background, and used a sliding window to search in the target candidate area until the accurate position of the target is determined.…”

Section: ⅱ Related Workmentioning

confidence: 99%

An Infrared Target Images Recognition and Processing Method Based on the Fuzzy Comprehensive Evaluation

Wu,

He,

Zhao

2024

IEEE Access

View full text Add to dashboard Cite

In response to the issues of relatively low image contrast, blurry boundaries of infrared small targets, and various environmental noise interferences in thermal imaging systems, resulting in reduced target recognition rates and tracking effectiveness. This paper proposes an infrared target image recognition algorithm based on fuzzy comprehensive assessment. Built upon infrared image sequences captured by an on-board thermal imager, the algorithm employs grayscale processing, neighborhood mean filtering, and second-order difference method to filter and enhance the infrared target images. This process improves the quality of infrared target images and enhances the contrast between targets and backgrounds. The algorithm utilizes the maximum interclass variance method for segmenting the infrared target images, combining it with the Canny operator to extract target edges, thereby obtaining clearer details of the infrared target edges. Finally, based on the extracted target feature information, the algorithm calculates the confidence of each marked region, establishes the membership function of target likelihood, employs a comprehensive weighted approach to construct the target confidence function, and compares the confidence of marked regions with that of the template image to achieve precise target recognition. Experimental validation against theoretical methods and comparisons with other approaches demonstrate the effectiveness and feasibility of the proposed method, providing theoretical support for reliable target tracking in electro-optical imaging systems.

show abstract

Section: ⅱ Related Workmentioning

confidence: 99%

An Infrared Target Images Recognition and Processing Method Based on the Fuzzy Comprehensive Evaluation

Wu,

He,

Zhao

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Yolact [5] is an efficient single-stage instance segmentation model that draws on the Yolo [22,23,24,25] series of object detection and is real-time. Compared with other instance segmentation models, Yolact focuses on speed, and the model has been greatly improved.…”

Section: Yolact Networkmentioning

confidence: 99%

A multiscale real-time instance segmentation method based oncleaning rubber ball images

Tian

Wang

et al. 2022

Preprint

View full text Add to dashboard Cite

The identification of wear rubber balls in the rubber ball cleaning device in heat exchange equipment directly affects the descaling efficiency. For the problem that the rubber ball image contains impurities and bubbles and the segmentation is low in real time, a multi-scale feature fusion real-time instance segmentation model based on the attention mechanism is proposed for the object segmentation of the rubber ball images. First, we introduce Pyramid Vision Transformer instead of convolution module in the backbone network, and use Spatial-reduction attention layer of transformer to enhance the feature extraction capability at different scales as well as spatial reduction to reduce the computational cost. Then improve the feature fusion module to fuse image features across scales, combined with an attention mechanism to enhance the output feature representation; secondly the prediction head separates the mask branches separately combined with dynamic convolution and depths to improve the accuracy of the mask coefficients. Through the validation of the produced rubber ball dataset, the Dice score, Jaccard coefficient, and mAP of the actual segmented region of this network with the glueball dataset improved by 4.4\%, 4.5\%, and 7.56\%, respectively, and the segmentation speed is also 35.1 fps. The proposed modules can work together to better handle object details and realize better segmentation performance.

show abstract

“…23 SqueezeNet uses a well-designed compression and re-expansion structure, MobileNet uses a more efficient deep separable convolution, and ShuffleNet proposes a channel scrubbing operation that reduces the computation of the model. Yujie et al 24 greatly reduced the parameters of the original network by introducing deeply separable convolution instead of ordinary convolution and replaced the original ReLU activation function with a lightweight activation function, which improved the detection accuracy of the model. Chen et al 25 proposed a structural innovation based on the YOLOv5-MobileNetv3Small network model for low-altitude images of drones.…”

Section: Introductionmentioning

confidence: 99%

Insulator defect detection based on improved you-only-look-once v4 in complex scenarios

Wang

Shao

et al. 2023

J. Electron. Imag.

View full text Add to dashboard Cite

As power grid structure becomes increasingly complex, the impact of the operation status of key substation equipment on the safe and stable operation of the power grid is gradually increasing. Aiming at the problems of dense targets and occlusion recognition difficulties in the complex background of substations, an improved you-only-look-once (YOLO) v4 insulator defect detection algorithm is proposed. A large number of insulator and defect datasets were collected from the Internet. We use MobileNet-v2 to replace the backbone network of YOLOv4, and in view of the problem that the accuracy of YOLOv4 target detection model is not high for small targets or dense targets, convolutional block attention module (CBAM) attention mechanism is added to the backbone network to enhance the feature extraction ability, and ultra-lightweight subspace attention module (ULSAM) is added to the detection head to learn the cross-channel information. Experiments based on InDataset show that the insulator defect detection algorithm based on deep learning proposed achieves a high detection accuracy and detection speed; the mean average precision of the improved YOLOv4 model is 87.48%, and the recall rate of insulator defects reaches 79.84%, which meets the requirements of robustness and accuracy of insulator defect detection in real scenarios.

show abstract

Lightweight target detection algorithm based on YOLOv4

Cited by 10 publications

References 17 publications

An Infrared Target Images Recognition and Processing Method Based on the Fuzzy Comprehensive Evaluation

An Infrared Target Images Recognition and Processing Method Based on the Fuzzy Comprehensive Evaluation

A multiscale real-time instance segmentation method based oncleaning rubber ball images

Insulator defect detection based on improved you-only-look-once v4 in complex scenarios

Contact Info

Product

Resources

About