In order to solve the current problems of insufficient detail extraction and poor visual effect after high magnification reconstruction of face images, a super-resolution method is proposed for single images of faces based on generative adversarial networks. Channel attention is added to the generative network to extract richer facial details, and the idea of iterative up and down sampling layers in the depth inverse projection network is borrowed to make the reconstructed image with good visual effect after high magnification. For the discriminator network, the normalization layer, which would destroy the image contrast, is removed. The experimental results show that the reconstructed images are more realistic and the visual effects are improved compared with Bicubic, SRCNN, LapSRN and SRGAN.
To reduce the casualties caused by workers not wearing safety measures in time at construction sites. For the generic target detection model with high complexity and large model, which cannot perform helmet wearing detection in real-time, a real-time helmet wearing detection algorithm based on YOLOv3 is proposed, with the mobile end network as the feature extractor of the proposed method. For helmet wearing detection is often done outdoors, so to filter the environmental noise. The channel attention module is introduced to optimize the feature extraction when multiplexing the multi-scale feature maps. Finally, to weaken the problem of inadequate gradient back propagation brought by the IOU function, the CIOU loss function is used to optimize the gradient back propagation. The experimental results show that the method in this paper can balance the accuracy and detection rate of the detection model with an average accuracy of 82.47% (mAP). Compared with the model of the original YOLOv3 network, the model size of this method is only 22.3% of the original model, and the detection rate is significantly improved compared with the original method and can meet the requirements of real-time helmet wear 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.