Authentication email has already been sent, please check your email box: and activate it as soon as possible.
You can login to My Profile and manage your email alerts.
If you haven’t received the email, please:
|
|
There are 67 papers published in subject: > since this site started. |
Select Subject |
Select/Unselect all | For Selected Papers |
Saved Papers
Please enter a name for this paper to be shown in your personalized Saved Papers list
|
1. End-to-End No-Reference Video Semantic Communication Quality Assessment via Deep Neural Networks | |||
Zhang Baiquan,Que Xirong | |||
Computer Science and Technology 12 January 2024 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Video semantic communication is developing rapidly nowadays, but traditional video quality assessment methods are not fully compatible with it. There is a lack of a no-reference video quality assessment method specifically designed for video semantic communication. In this paper, we propose a new end-to-end no-reference video semantic communication quality assessment using deep neural networks. Our model, which is implemented based on video semantic communication, it creatively uses both common and individual features extracted from videos through semantic communication for video quality assessment. Our model adopts a multi-task DNN framework, which assesses the quality of both common and individual features, and finally combines both to obtain the final video quality prediction score. Experimental results show that our assessment model outperforms other no-reference video quality assessment methods and is more suitable for semantic communication. | |||
TO cite this article:Zhang Baiquan,Que Xirong. End-to-End No-Reference Video Semantic Communication Quality Assessment via Deep Neural Networks[OL].[12 January 2024] http://en.paper.edu.cn/en_releasepaper/content/4761748 |
2. Light field Stitching via 4D Homography | |||
DAI Yi-chen,CAI Min-jie | |||
Computer Science and Technology 14 May 2023 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:The problem of the light field (LF) stitching aims to align two 4D LFs seamlessly. However, the prior works use conventional $3\times3$ homography to draw the 2D relation and ignore the depth information, leading to two main disadvantages, namely, significant stitching artifacts in the general scene and failure to produce stitched depth map. This paper tackles these challenges by proposing a $4\times4$ homography that analytically and globally describes the relationship between two LFs under pure rotation. Besides, we also present a novel linear solver called 4ry, which can estimate the 4D homography by giving four 4D LF feature correspondences. Extensive synthetic and real data experiments demonstrate that the proposed method outperforms state-of-the-art approaches in LF stitching qualitatively and quantitatively. More importantly, the output of our method is still an LF that retains the nature of LF, such as refocusing, viewpoint shifting, and depth estimation. | |||
TO cite this article:DAI Yi-chen,CAI Min-jie. Light field Stitching via 4D Homography[OL].[14 May 2023] http://en.paper.edu.cn/en_releasepaper/content/4760779 |
3. Moving Object Detection Algorithm Based on Dynamic Vision Sensor | |||
SUN Xue,SUN Xue,LIU Dengfeng,LIU Dengfeng | |||
Computer Science and Technology 11 April 2023 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Target detection and tracking mainly use RGB camera and deep learning algorithm. The information in the image is complex and redundant so the computing consumes a lot of resources. To solve the above problems, this paper proposes an improved spectral clustering algorithm to detect moving object. The algorithm is based on event data generated by dynamic vision sensors. In this paper, the cosine - Manhattan fusion distance is used to obtain a more accurate similarity matrix. The clustering results of some data are used to guide other data to speed up operation. The number of clusters is set adaptively to avoid the subjective influence of human beings. The results show that the accuracy of the improved algorithm on multiple data sets is more than 80%, and the time is significantly shortened. Spectral clustering algorithm based on dynamic vision sensor has great application potential in dealing with multi-target motion problems. | |||
TO cite this article:SUN Xue,SUN Xue,LIU Dengfeng, et al. Moving Object Detection Algorithm Based on Dynamic Vision Sensor[OL].[11 April 2023] http://en.paper.edu.cn/en_releasepaper/content/4760291 |
4. Attention-augmented domain-adaptive semantic segmentation of remote sensing images | |||
XUE Bingjie,CAI Minjie | |||
Computer Science and Technology 28 April 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:In order to achieve semantic segmentation of unlabeled images using labeled remote sensing image datasets, an unsupervised domain adaptive method is proposed in this paper to solve this problem. Firstly, a self-training network for semantic segmentation of remote sensing images based on pseudo labels selection is established. Secondly, in view of the shortcoming that convolution network model can\'t effectively utilize long-range dependence information, attention module is added to the segmentation network part to effectively utilize the relationship between image dependence information and channels. Finally, in order to learn domain invariant features, a domain classifier is added on the basis of the self-training network, which ensures that the domain classifier can\'t distinguish which domain the sample comes from while ensuring the segmentation performance, thus reducing the domain offset between the source domain and the target domain. Experiments on LoveDA dataset show that this method is superior to the mainstream unsupervised domain adaptive method, and achieves good results in unsupervised domain adaptive semantic segmentation of remote sensing images in rural and urban areas. Exploring the deep transfer learning method on this dataset will be a meaningful way to promote large-scale land-cover mapping. | |||
TO cite this article:XUE Bingjie,CAI Minjie. Attention-augmented domain-adaptive semantic segmentation of remote sensing images[OL].[28 April 2022] http://en.paper.edu.cn/en_releasepaper/content/4757582 |
5. Design and Implementation of Handwritten Chinese Characters Recognition Based on Deep Learning | |||
Jingyi Shen | |||
Computer Science and Technology 24 March 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Nowadays, people are increasingly using electronic formats to store data. In order to convert the table in paper into digital information, this project implements a handwritten Chinese character recognition system for table pictures, which releases the pressure of manual workload of typing handwritten Chinese character information. This project pre-processes the scanned images, extracts the handwritten text information from the images . Then segment the cell in the table and cuts the individual Chinese characters. In order to improve the accuracy of character cutting, this project adopts the combination of vertical projection and aspect ratio of character to determine the cutting position. The ResNext50 model is used for model training, and the two models are trained to recognize numbers, letters and handwritten Chinese characters respectively. The accuracy of the Chinese character recognition model is more than 90%, and that of the number and letter recognition model is 98%. Based on the contents filled in the table, the list of proper nouns is used to correct the recognition results and improve the accuracy. By calculating Levenshtein distance find the specific nouns with the highest similarity. The method proposed in this paper effectively complete the separation and recognition of handwritten Chinese characters in the table image. | |||
TO cite this article:Jingyi Shen. Design and Implementation of Handwritten Chinese Characters Recognition Based on Deep Learning[OL].[24 March 2022] http://en.paper.edu.cn/en_releasepaper/content/4756976 |
6. Fusion algorithm of infrared and visible images based on LatLRR and image pyramid | |||
LI Ning,LI Liqun | |||
Computer Science and Technology 23 March 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Aiming at the loss of high-frequency information components of visible light and infrared images fused by Latent Low-Rank Representation (LatLRR) algorithm, this paper proposes a multi-level image fusion algorithm based on the combination of LatLRR and Gauss-Laplace Pyramid. Firstly, the image is decomposed into Gaussian pyramid and Laplace pyramid, then the highest level image of Gaussian pyramid and each layer image of Laplace pyramid are decomposed into low-rank part and Significant part, meanwhile the low-rank part and significant part of infrared and visible image pyramid are blended, then the fused low-rank part and significant part are fused at all levels, and finally the image is restored by Gaussian-Laplace pyramid. The algorithm has been tested on the public data set, this method retains more high-frequency information components compared with LatLRR algorithm. | |||
TO cite this article:LI Ning,LI Liqun. Fusion algorithm of infrared and visible images based on LatLRR and image pyramid[OL].[23 March 2022] http://en.paper.edu.cn/en_releasepaper/content/4757034 |
7. THMobile : An Improved Network For Garbage Classification Based on MobileNet | |||
Zhou Jialan,Bian Jiali | |||
Computer Science and Technology 22 February 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:With the rapid development of deep learning, more and more image recognition models are applied to daily life. For the current neural network model, the recognition accuracy of large model is higher and higher, but the more resources are needed. The lightweight of neural network model is more conducive to the application in life. In this paper, a THMobile model with smaller size and higher accuracy is proposed based on MobileNet. On the self-made garbage dataset, the classification accuracy of it reaches 91.2%, obtaining better performance than MobileNet. And it also performs better on CIFAR-10 than MobileNet. | |||
TO cite this article:Zhou Jialan,Bian Jiali. THMobile : An Improved Network For Garbage Classification Based on MobileNet[OL].[22 February 2022] http://en.paper.edu.cn/en_releasepaper/content/4756338 |
8. Efficient video transmission scheme based on deep compressed sensing | |||
YANG Zihang,LI Lixiang | |||
Computer Science and Technology 21 February 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:With the rapid development of wireless communication technology and the popularity of mobile video input devices, wireless video transmission technology has been widely used in intelligent transportation, intelligent industry, intelligent security and other fields, which also brings many security problems. In this paper, a secure, fast and efficient video transmission scheme (SFE-VTS) for video sensor network is designed by combining deep compressed sensing and adaptive video codec technology. In the coding end, the new adaptive selection algorithm of sampling position and sampling rate of video frame block reduces the amount of data transmitted by redundant information and solves the problem of fluctuation of recovery quality between adjacent non-key frames. In the decoding side, the recovery algorithm based on deep learning can reconstruct the video frame quickly. In this paper, the proposed scheme can effectively solve the traditional video transmission scheme, encoding end takes up too much resource and transport process safety is not high. In addition, compared with some video transmission schemes based on traditional compressed sensing, the recovery effect and efficiency are higher. | |||
TO cite this article:YANG Zihang,LI Lixiang. Efficient video transmission scheme based on deep compressed sensing[OL].[21 February 2022] http://en.paper.edu.cn/en_releasepaper/content/4756385 |
9. Compressive Hyperspectral Video Reconstruction Via Multitask Nonparametric Bayesian Learning | |||
Yang Man,Gao Zhanchun | |||
Computer Science and Technology 09 February 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Compressive hyperspectral video reconstruction (CHVR) extends snapshot spectral imaging into the temporal dimension, which allows fast hyerspectral observation of dynamic scenes. This paper proposes a multitask learning method for CHVR under the blind compressive sensing framework, characterized by joint inference of the representation atoms and the corresponding coefficients, directly from the compressive measurements. Defining the compressive reconstruction of each frame as a single task, our method employs a common dictionary shared by all tasks, which significantly alleviates the data paucity problem. The complete inference process is fulfilled via a Bayesian nonparametric estimation strategy, which contributes three advantages: reliable generalizability, parameter-tuning free and automatic determination of the model complexity. Simulation results demonstrate the efficacy of the proposed approach. | |||
TO cite this article:Yang Man,Gao Zhanchun. Compressive Hyperspectral Video Reconstruction Via Multitask Nonparametric Bayesian Learning[OL].[ 9 February 2022] http://en.paper.edu.cn/en_releasepaper/content/4756225 |
10. A region-based error concealment method for livestream videos | |||
Wang Mengyuan,Wang Jing | |||
Computer Science and Technology 29 January 2022 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:This paper proposes a video error concealment method for the reconstruction of large size corrupted regions of livestream videos. The method adopts different spatial and temporal error concealment methods according to the characteristics of livestream frames and can keep a well balance on time-cost and performance. First, the method named motion vector estimation is performed for small motion regions. Then it uses the improved adaptive homography transform to reconstruct unknown regions in each frame. Finally, a spatial error concealment method known as bilinear interpolation is performed to conceal the macroblocks with poor effect after above two steps. Experiment results show that the proposal has the same performance as the baseline paper for livestream videos, and it is less time-consuming which can be applied in real livestream applications. | |||
TO cite this article:Wang Mengyuan,Wang Jing. A region-based error concealment method for livestream videos[OL].[29 January 2022] http://en.paper.edu.cn/en_releasepaper/content/4756191 |
Select/Unselect all | For Selected Papers |
Saved Papers
Please enter a name for this paper to be shown in your personalized Saved Papers list
|
About Sciencepaper Online | Privacy Policy | Terms & Conditions | Contact Us
© 2003-2012 Sciencepaper Online. unless otherwise stated