A better backbone network usually benefits the performance of various computer visionapplications. This paper aims to introduce an effective solution for infection percentage estimation of COVID-19 for the computed t...
详细信息
ISBN:
(纸本)9783031133244;9783031133237
A better backbone network usually benefits the performance of various computer visionapplications. This paper aims to introduce an effective solution for infection percentage estimation of COVID-19 for the computed tomography (CT) scans. We first adopt the state-of-the-art backbone, Hierarchical Visual Transformer, as the backbone to extract the effective and semantic feature representation from the CT scans. Then, the non-linear classification and the regression heads are proposed to estimate the infection scores of COVID-19 symptoms of CT scans with the GELU activation function. We claim that multi-tasking learning is beneficial for better feature representation learning for the infection score prediction. Moreover, the maximum-rectangle cropping strategy is also proposed to obtain the region of interest (ROI) to boost the effectiveness of the infection percentage estimation of COVID-19. The experiments demonstrated that the proposed method is effective and efficient.
With the vigorous development of social economy and science and technology, the protection and inheritance of folk art are facing great challenges. Folk art has a long history. In the context of in-depth learning, the...
详细信息
ISBN:
(纸本)9783031243660;9783031243677
With the vigorous development of social economy and science and technology, the protection and inheritance of folk art are facing great challenges. Folk art has a long history. In the context of in-depth learning, the protection and development of folk art has also reached a new level. Using advanced and new technology to serve folk art is one of the effective ways to revitalize and develop folk art. This paper analyzes the virtual reality technology, and puts forward to inherit and protect the folk art by using the visual direction optimization method in the virtual reality technology and its characteristics of perception, existence and interactive operation.
With the continuous development of furniture design, the machining accuracy and surface quality of die steel have been paid more and more attention. The traditional grinding process has problems such as low efficiency...
详细信息
With the continuous development of furniture design, the machining accuracy and surface quality of die steel have been paid more and more attention. The traditional grinding process has problems such as low efficiency and unstable quality, so it is urgent to introduce advanced technical means to improve the intelligent level of the processing process. This study aims to explore the application of the die steel grinding process based on machinevision and wireless sensor network equipment in furniture design, and improve the efficiency and quality of the grinding process through real-time monitoring and data analysis. A grinding monitoring platform integrating machinevision system and wireless sensor network was developed. A machinevision system is used to capture critical image data during the grinding process in real time, while a wireless sensor network is used to collect and transmit grinding parameters, including temperature, vibration and acoustic emission signals. By analyzing the acquired data, the optimized grinding parameters and control strategy are worked out. The experimental results show that the grinding process using machinevision and wireless sensor network has improved the relevant parameters compared with the traditional methods. The real-time monitoring capability of the system significantly reduces the failure rate during grinding and provides a more stable and reliable die steel processing solution for furniture design.
In real-world applications, images and videos used in computer vision algorithms are often distorted due, e.g., to compression and transmission. As a result, they may lose relevant information content, or they may dev...
详细信息
ISBN:
(纸本)9789082797091
In real-world applications, images and videos used in computer vision algorithms are often distorted due, e.g., to compression and transmission. As a result, they may lose relevant information content, or they may deviate significantly from the original data distribution used to train the machine task, rendering the visual content practically useless with respect to its initial purpose. Evaluating the utility of an image for machine tasks has received little attention so far in the literature. This concept of utility is substantially different from the visual quality typically used in image/video compression, as the latter is related to the perception of the human visual system. In this paper, we propose a definition of utility as the degree of confidence by which a machine task is able to take a decision. In this context, we propose a full-reference utility loss measure: we assume that the decision on the pristine image is correct (reference), and we measure the utility loss as the confidence reduction in the decision due to a noisy input with respect to this reference. We apply this general definition on two specific tasks, classification and object detection, and we study practical solutions to predict utility, as well as the ability of our utility measure to generalize across tasks.
Gradient computing is a low-level technology widely used in imageprocessing. For large gradient magnitude, the pixel value in the field changes a lot, and for small gradient magnitude the pixel in the domain changes ...
详细信息
ISBN:
(纸本)9781665464680
Gradient computing is a low-level technology widely used in imageprocessing. For large gradient magnitude, the pixel value in the field changes a lot, and for small gradient magnitude the pixel in the domain changes little. This is the basis of classical edge extraction algorithms, but it is often necessary to manually set thresholds to differentiate. This paper innovatively brings out the concept of omnidirectional gradient, which uses flexible convolution kernel radius and special law to calculate, and omnidirectional gradient pays more attention to gradient direction and analyzes the relationship and change of the gradient direction with different kernel radius. We present here an algorithm for stylized edge extraction based on omnidirectional gradient, overcoming the drawback of classical edge extraction algorithms that require manual thresholding. Experimental results show that the proposed method outperforms the classical edge extraction methods in terms of adaptive, consistent, and visually friendlier features for infrared imaging. In addition, the algorithm is fast and efficient, its result can be used as real-time input for subsequent applications.
Face recognition has become an advanced area in the field of recognition and has various applications especially in biometrics, forensic investigations, smart advertising, national database for identity cards etc. Var...
详细信息
The quality control of metal implants is of utmost importance in the health sector to ensure patient safety and optimal performance. In this study, we present a novel machinevision-based nondestructive inspection (ND...
详细信息
In the past decade, deep neural networks have achieved significant progress in point cloud learning. However, collecting large-scale precisely-annotated point clouds is extremely laborious and expensive, which hinders...
详细信息
In the past decade, deep neural networks have achieved significant progress in point cloud learning. However, collecting large-scale precisely-annotated point clouds is extremely laborious and expensive, which hinders the scalability of existing point cloud datasets and poses a bottleneck for efficient exploration of point cloud data in various tasks and applications. Label-efficient learning offers a promising solution by enabling effective deep network training with much-reduced annotation efforts. This paper presents the first comprehensive survey of label-efficient learning of point clouds. We address three critical questions in this emerging research field: i) the importance and urgency of label-efficient learning in point cloud processing, ii) the subfields it encompasses, and iii) the progress achieved in this area. To this end, we propose a taxonomy that organizes label-efficient learning methods based on the data prerequisites provided by different types of labels. We categorize four typical label-efficient learning approaches that significantly reduce point cloud annotation efforts: data augmentation, domain transfer learning, weakly-supervised learning, and pretrained foundation models. For each approach, we outline the problem setup and provide an extensive literature review that showcases relevant progress and challenges. Finally, we share our views on the current research challenges and potential future directions.
Monitoring and maintenance of water resources projects is essential to ensure project safety and environmental protection. Traditional monitoring methods often rely on manual inspections and sensor data, but these met...
详细信息
Some of these so-called AIoT applications include intelligent imageprocessing in smart factories to monitor machinery conditions and control raw material inventory, identifying abnormalities in medical images, and au...
详细信息
Some of these so-called AIoT applications include intelligent imageprocessing in smart factories to monitor machinery conditions and control raw material inventory, identifying abnormalities in medical images, and automatic real-time scanning and recognition of license plates in traffic to locate stolen cars.
The topics covered in this special issue include (i) intelligent imageprocessingapplications and services to fulfill the real-time processing and performance demands, (ii) real-time deep learning and machine learning solutions to improve computational speed and increase recognition rates at network edges, (iii) new frameworks to optimize real-time AIoT imageprocessing, and (iv) combining intelligent real-time imageprocessing with edge computing, fog computing, and relevant techniques to balance the computational workloads between IoT devices and the server side.
Fan and Guan [1] have developed a deep face verification framework based on SIFT (scale invariant feature transform) and CNN (convolutional neural network) methods.
暂无评论