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 72 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. Research on entity extraction of demand for industry-university-research projects based on BERT-BiLSTM-CRF | |||
ZHANG Zhiqing,TAO Zekui | |||
Computer Science and Technology 04 January 2024 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:In order to realize the effective docking between schools and enterprises in industry-university-research cooperation projects and accurately obtain the technical needs of enterprises, based on the characteristics of concise and diversified texts of industry-university-research projects, a Chinese named entity extraction method based on BERT+BiLSTM+CRF model was proposed. Firstly, the BERT model is used to encode the input text, then the BiLSTM model is used to model the context to capture more comprehensive context information, and finally the label decoding is carried out through the CRF layer to obtain the optimal entity annotation results. Experimental results show that the proposed method is effective and feasible, and the extraction effect is better than that of the traditional method, which provides the possibility to solve the difficulties of technical demand information extraction, such as polysemy and language variants. | |||
TO cite this article:ZHANG Zhiqing,TAO Zekui. Research on entity extraction of demand for industry-university-research projects based on BERT-BiLSTM-CRF[OL].[ 4 January 2024] http://en.paper.edu.cn/en_releasepaper/content/4761868 |
2. Exploring Developer Social Networks: Unveiling the Impact on New Commit Activity in GitHub | |||
WAN Zhi-Jie, *WANG Yi | |||
Computer Science and Technology 01 December 2023 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Open source development (OSS) is a platform for developers to work collaboratively to finish a common project, which gives birth to the developer social network (DSN). Such DSNs provide valuable insights into knowledge flow, coordination effectiveness, innovation, and the diffusion of practices and technologies. Besides, the network structure, characterized by size, density, bridge, and degree centrality, also influences team cohesion, coordination efficiency, and the emergence of specialized expertise. We visualize 80 DSNs constructed from the empirical data of 80 popular projects in \textsc{GitHub}, identifying these DSNs' characteristics and employing a regression model to estimate the correlation between DSNs' properties and the average number of monthly new commits ($\overline{\textbf{NewC}}$). Our analysis reveals three key findings: (1) the effects of DSN size and DSN bridge are positively correlated with $\overline{\textbf{NewC}}$; (2) the effects of DSN density exhibit a weak negative correlation with $\overline{\textbf{NewC}}$; and (3) no relationship exists between DSN average degree centrality and $\overline{\textbf{NewC}}$. These results provide an integrated view of DSNs' structural characteristics and can inform software managers to enhance project management, team collaboration and software development outcomes. | |||
TO cite this article:WAN Zhi-Jie, *WANG Yi. Exploring Developer Social Networks: Unveiling the Impact on New Commit Activity in GitHub[OL].[ 1 December 2023] http://en.paper.edu.cn/en_releasepaper/content/4761606 |
3. Deep Parallel Neural Network for Event Temporal Relation Classification | |||
Sun Chang,Gao Sheng | |||
Computer Science and Technology 13 April 2021 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Temporal relation classification is an active natural language understanding research field. Partly because of the lack of high-quality datasets, most of the methods proposed in other papers do not use neural network or they only use simple neural network with external knowledge. In this task, we use Convolution Neural Network (CNN) to discover the complex interaction between events and extract time-related keyword information, and we use Long Short-Term Memory (LSTM) to capture temporal context information in sentences, and then connect the two kinds of information for temporal relation classification. We don\'t use any external knowledge, including dependency trees. In the latest MATRES dataset, the performance of our model is better than the state-of-the-art result. | |||
TO cite this article:Sun Chang,Gao Sheng. Deep Parallel Neural Network for Event Temporal Relation Classification[OL].[13 April 2021] http://en.paper.edu.cn/en_releasepaper/content/4754461 |
4. A novel key frame selection method for aerial image stitching by integrating navigation information and trusted key points | |||
Zheng Yongji,Wang Guoyou | |||
Computer Science and Technology 01 April 2021 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Fast and high-precision video based image stitching plays an important role in many machine vision applications, such as UAV mapping and reconnaissance. Due to the large number of frames and high redundancy of video sequence, image stitching is very time-consuming. Therefore, from the perspective of reducing the number of redundant frames, this paper proposes a novel video sequence key frame selection method based on rough camera external parameters and key point fusion, which selects the appropriate key frames by optimizing the overlap rate and the number of reliable key points between two adjacent key frames. The algorithm not only greatly reduces the number of key frames, but also ensures the reliable video mosaic. The experimental results on Bu S\' videos show that our method can reduce the number of key frames by 92%. In addition, compared with the key frame selection method based only on navigation information, this method also overcomes the problem of missing stitched images caused by insufficient key points in overlapping regions. | |||
TO cite this article:Zheng Yongji,Wang Guoyou. A novel key frame selection method for aerial image stitching by integrating navigation information and trusted key points[OL].[ 1 April 2021] http://en.paper.edu.cn/en_releasepaper/content/4754321 |
5. Head Pose Estimation Based on Dlib and Savitzky-Golay Smoothing Algorithm | |||
LU Xiaoning,LIU Wen | |||
Computer Science and Technology 18 January 2021 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:In this paper, We apply the Savitsky-Golay filter algorithm to head pose estimation. Firstly, the Dlib algorithm is to detect key points of the corresponding face in the video. Then, the function solvepnp built in opencv is to estimate the pose. Finally, through MATLAB simulation, we can find that the Savitsky-Golay filter algorithm can filter the noisiness in the observed data to obtain a smoother and more accurate change trajectory of head pose. | |||
TO cite this article:LU Xiaoning,LIU Wen. Head Pose Estimation Based on Dlib and Savitzky-Golay Smoothing Algorithm[OL].[18 January 2021] http://en.paper.edu.cn/en_releasepaper/content/4753447 |
6. Adaptive packet scheduling algorithm for time-sensitive service | |||
ZHANG Zhenjie,Jianfeng Guan | |||
Computer Science and Technology 18 December 2020 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:With the rapid development of mobile Internet and Internet of things, a series of time sensitive services such as video conferencing, cloud games, AR / VR have emerged. In order to meet the above time sensitive services, IETF proposed Deterministic Network(DetNet) architecture, which provides an ideal deterministic delay through clock synchronization, zero congestion loss and other mechanisms. At the same time, Tsinghua University proposed Deadline-aware transport protocol(DTP), hoping to specify the deadline in the application layer, and then meet the requirements in the transport layer. These strategies and ideas for time delay sensitive services are worth learning from. However, the current packet transmission mechanisms are all rule-based and relatively static strategies, which can’t meet the demand of time sensitive service in dynamic network. Therefore, based on the idea of DTP and reinforcement learning, this paper proposes an algorithm to dynamically adjust the transmission priority. More specifically, we design the reward and penalty function according to the requirements of DTP protocol, and design the algorithm of congestion control and packet scheduling in the transport layer. We consider not only the priority but also the service deadline. Comprehensive experiments show that compared to traditional packet scheduling strategies, our algorithm performs better in the transmission of time-sensitive services | |||
TO cite this article:ZHANG Zhenjie,Jianfeng Guan. Adaptive packet scheduling algorithm for time-sensitive service[OL].[18 December 2020] http://en.paper.edu.cn/en_releasepaper/content/4753219 |
7. Domain adaptive image retrieval based on region of interest | |||
Zhao Zhen,Ai Xinbo | |||
Computer Science and Technology 03 March 2020 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Recently, the explosive growth of image data, how to retrieve effective images has become an urgent problem. However, image retrieval often faces the following problems.In the current image retrieval model, the information of local area of interest is less considered. When images exist in two different domain distributions, cross-domain retrieval cannot be performed effectively.In view of the current existence of the above problems, this paper put forward based on the interested region of domain adaptive image retrieval methods, including the interest of the target detection technology of image area, the interference of background information filter is invalid, feature fusion method for multi-objective regional characteristics of effective at the same time to join the different domain image domain structure, realization of cross-domain retrieval.In this paper, we evaluated the effectiveness of our method on the PASCAL VOC dataset. | |||
TO cite this article:Zhao Zhen,Ai Xinbo. Domain adaptive image retrieval based on region of interest[OL].[ 3 March 2020] http://en.paper.edu.cn/en_releasepaper/content/4750996 |
8. Conversational Recommendation System based on Sentiment Analysis | |||
LI Xinsheng,LI Jian | |||
Computer Science and Technology 25 February 2020 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:The combination of the recommender system and dialogue system which called the conversational recommendation system is a growing interest. Tosolve the problem that it is difficult to obtain users' tastes in conversational recommendation systems. A sentiment analysis method is proposed in our conversational recommendation model to get user preferences. A sentiment analysis dataset is created and the model uses a sentiment analysis approach to obtain a movie seeker\'s preferences and make a recommendation. Experimentresults show that our sentiment analysis model yields a better performance of 0.8362(F1 score) than the baseline(0.7802) and other models. Thus, the movie recommended by our system can meet the needs of users better. | |||
TO cite this article:LI Xinsheng,LI Jian. Conversational Recommendation System based on Sentiment Analysis[OL].[25 February 2020] http://en.paper.edu.cn/en_releasepaper/content/4750892 |
9. An Improved Visual-Inertial Odometry Based on Self-Adaptive Attention-Anticipation Feature Selector | |||
Ruan Wenlong,Wang Jing | |||
Computer Science and Technology 22 November 2019 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Visual inertia odometers have achieved great success with the development of robot vision. However, it remains a challenging problem to achieve robust and efficient pose estimation on low-power platforms such as smartphones. This paper proposes a new visual inertial odometer scheme for low-power platforms, named visual inertial odometer based on adaptive attention-anticipation mechanism, which adds visual information to the VINS-based visual inertial odometer. The attention distribution module and the motion information forward anticipation module are controlled by the adaptive adjustment module to reduce the system operation load and improve the system tracking accuracy. We contribute in the following three aspects: 1) A attention mechanism for visual inertia history is proposed, which provides visual attention distribution for system radical motion in complex space environment, and extracts vision with high weight on system influence. Feature tracking; 2) A visual feature screening mechanism based on motion prediction is proposed to filter the visual features that will escape the camera perspective in advance; 3) use the adaptive adjustment module for front-end control and efficiently allocate restricted computing resources. Our approach achieves advanced estimation performance on the Euroc MAV datasets. | |||
TO cite this article:Ruan Wenlong,Wang Jing. An Improved Visual-Inertial Odometry Based on Self-Adaptive Attention-Anticipation Feature Selector[OL].[22 November 2019] http://en.paper.edu.cn/en_releasepaper/content/4750006 |
10. A Lightened Sphereface for Face Recognition | |||
ZHOU Xinjie,Zhenxue Chen,WANG Mengxue | |||
Computer Science and Technology 25 June 2019 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Convolution neural networks (CNN) have significantly promoted the development of face recognition technology. In order to achieve ultimate accuracy, CNN models tend to be deeper or multiple local facial patch ensembles, resulting in excessive amounts of calculation. This paper addresses these deep face recognition (FR) problems and studies a lightened deep learning framework under an open-set protocol to achieve a good classification effect and streamline the model itself. To this end, we improve the Sphereface that enables deep network to learn angularly discriminative features faster and more effectively. First, global average pooling (GAP) is introduced to replace the original fully connected layer, which greatly reduces the size of the model. Compared to the widely used fully connected layer, GAP can reduce the number of parameters and avoid overfitting. Then Network in Network (NIN) layers are added between convolution layers. These models are trained on the CASIA-WebFace dataset and evaluated on the LFW and YTF datasets, which show the superiority of lightened SphereFace (L-SphereFace) in FR tasks. At the same time, computational cost is reduced by over nine times in comparison with the released SphereFace model. The size of the model is also close to the original half. | |||
TO cite this article:ZHOU Xinjie,Zhenxue Chen,WANG Mengxue. A Lightened Sphereface for Face Recognition[OL].[25 June 2019] http://en.paper.edu.cn/en_releasepaper/content/4749157 |
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