In supervised learning, standard algorithms are mostly designed to deal with balanced data classes, but it is inevitable to encounter imbalanced data classes in some situations. How to learn from imbalanced data is still a challenging problem. A common approach is to generate artificial data or duplicate existing classes to balance the class distribution. Many traditional algorithms can handle imbalanced classes, but they often fail to enhance enough features and tend to produce unnecessary noise points. In this paper, we propose an improved network based on generative adversarial networks (GANs) and improved K-means Smote oversampling method to oversample the data. This method replaces the noise input of GANs with the oversampled results, and then generates new oversampled data through adversarial networks. By conducting experiments on six datasets, we show that this method can effectively improve the classification results of classifiers on oversampled data.
A lightweight detection algorithm based on YOLOv5 is proposed to solve the problems that the target detection algorithm has many network parameters and the large size of the model is not conducive to the actual deployment. Firstly, the depth separable convolution was introduced to reconstruct the backbone network to reduce the network parameters. Then, the residual element was redesigned by ghost convolution and integrated into the Neck end to further reduce the model volume. Finally, the detection accuracy and recognition effect of the algorithm were improved by modifying the detection layer. When the recognition accuracy of the improved lightweight algorithm is reduced by 3.4%, the model volume is reduced by 75%, and the number of network parameters is reduced by 81%. The results show that the algorithm is able to maintain a high recognition accuracy, and the volume and parameters are greatly reduced. It is suitable for embedded and other system resource limited scenarios, and has a practical application prospect.
Complex environments, such as dense personnel and background interference, affect the detection accuracy of whether personnel wear helmets. To solve this problem, a new detection algorithm for helmets based on YOLOv5 is studied in this paper. Firstly, RepVGGblock is used to replace the common convolution in the network to effectively utilize the computing power of GPU. Secondly, the efficient channel attention mechanism is incorporated into the C3 module of the backbone network to enhance the feature identification ability of the backbone network for the hard hat. Then, the boundary box regression function is changed to SIoU to redefine the distance loss and improve the regression accuracy. Finally, the detection performance of multi-scale targets is improved by adding detection layers. The test results on the self-made safety helmet data set show that the average accuracy of the improved YOLOv5 model reaches 94.6%, which is 4.1% higher than that of the original model, and can meet the requirements of target detection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.