检索结果-内蒙古大学图书馆

Teachers cooperation:team-knowledge distillation for multiple cross-domain few-shot learning

Frontiers of Computer science 2023年第2期17卷 91-99页

作者： Zhong JI Jingwei NI Xiyao LIU Yanwei PANG School of Electrical and Information Engineering Tianjin UniversityTianjin 300072China Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin 300308China Tianjin Key Laboratory of Brain-Inspired Intelligence Technology Tianjin 300072China

Although few-shot learning(FSL)has achieved great progress,it is still an enormous challenge especially when the source and target set are from different domains,which is also known as cross-domain few-shot learning(CD-FSL).Utilizing more source domain data is an effective way to improve the performance of ***,knowledge from different source domains may entangle and confuse with each other,which hurts the performance on the target ***,we propose team-knowledge distllation networks(TKD-Net)to tackle this problem,which explores a strategy to help the cooperation of multiple ***,we distill knowledge from the cooperation of teacher networks to a single student network in a meta-learning *** incorporates task-oriented knowledge distillation and multiple cooperation among teachers to train an efficient student with better generalization ability on unseen ***,our TKD-Net employs both response-based knowledge and relation-based knowledge to transfer more comprehensive and effective *** experimental results on four fine-grained datasets have demonstrated the effectiveness and superiority of our proposed TKD-Net approach.

关键词： cross-domain few-shot learning meta-learning knowledge distillation multiple teachers

来源：评论

学校读者我要写书评

暂无评论

Model-based deep learning for fiber bundle infrared image restoration

引用

Defence technology（防务技术） 2023年第9期27卷 38-45页

作者： Bo-wen Wang Le Li Hai-bo Yang Jia-xin Chen Yu-hai Li Qian Chen Chao Zuo Jiangsu Key Laboratory of Spectral Imaging and Intelligent Sense Nanjing University of Science and TechnologyNanjingJiangsu Province210094China Smart Computational Imaging(SCI)Laboratory Nanjing University of Science and TechnologyNanjingJiangsu Province210094China Science and Technology on Electro-Optical Information Security Control Lab Tianjin300308China

As the representative of flexibility in optical imaging media,in recent years,fiber bundles have emerged as a promising architecture in the development of compact visual *** to tackling the problems of universal honeycomb artifacts and low signal-to-noise ratio(SNR)imaging in fiber bundles,the iterative super-resolution reconstruction network based on a physical model is *** the constraint of solving the two subproblems of data fidelity and prior regularization term alternately,the network can efficiently“regenerate”the lost spatial resolution with deep *** building and calibrating a dual-path imaging system,the real-world dataset where paired low-resolution(LR)-high-resolution(HR)images on the same scene can be generated *** results on both the United States Air Force(USAF)resolution target and complex target objects demonstrate that the algorithm can restore high-contrast images without pixilated *** the basis of super-resolution reconstruction,compound eye image composition based on fiber bundle is also embedded in this paper for the actual imaging *** proposed work is the first to apply a physical model-based deep learning network to fiber bundle imaging in the infrared band,effectively promoting the engineering application of thermal radiation detection.

关键词： Fiber bundle Deep learning Infrared imaging Image restoration

来源：评论

学校读者我要写书评

暂无评论

Scale Normalization and Lightweight FPN for Small Object Detection in Aerial Images 3

Scale Normalization and Lightweight FPN for Small Object Det...

引用

3rd International Conference on Computer Graphics, Image, and Virtualization, ICCGIV 2023

作者： Wei, Haokun Liu, Jingyi Chen, Jinyong Sun, Yuxin Chu, Boce Zhu, Jin Key Laboratory of Aerospace Information Applications of CETC Shijiazhaung China Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin China

ISBN: (纸本)9781510671720

The detection of small objects by oriented bounding box in aerial images is a recent hot topic. However, since the aerial images are not collected at the same height, the Ground Sample Distance (GSD) is different for each image, so that small objects are easily overlooked. Existing algorithms are designed for multi-scale object detection, and feature fusion is time-consuming, resulting in a large amount of model parameters that is not easy to deploy on embedded devices. We propose three methods to address the above problems. First, we scale the collected aerial images to the same scale according to the GSD value. Second, we change the structure of Feature Pyramid Network (FPN) and only keep the necessary low-level feature maps. Finally, we rescale the anchor for the specific scene. We validate our proposed method on the DOTA dataset. The results show that the modified model using our method can identify more small-scale objects, and the maximum number of model parameters can be reduced by 2.7%, the inference speed can be increased by 13.24%, and the model size was reduced by up to 28% when the detection accuracy is the same as the original algorithm. © 2023 SPIE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

6G Network Traffic Intrusion Detection Using Multiresolution Auto-encoder and Feature Matching Discriminator 5th

6G Network Traffic Intrusion Detection Using Multiresolution...

引用

5th EAI International Conference on 6G for Future Wireless Networks, 6GN 2022

作者： Li, Yuhai Sun, Yuxin He, Dong Xi, Liang Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin300300 China School of Computer Science and Technology Harbin University of Science and Technology Harbin150080 China

ISBN: (纸本)9783031360107

With the development of 6G technology, security and privacy have become extremely important in the face of larger network traffic bandwidth. An effective intrusion detection system can deal with the network attacks. Deep learning has been developed in the field of intrusion detection, which can identify normal and abnormal traffic. However, existing methods cannot guarantee good performance in accuracy and efficiency. In this paper, based on the autoencoder and generative adversarial network, the multiresolution autoencoder is adopted in the network traffic feature extraction, which can obtain different encoding lengths and guarantee better data reconstruction. In addition, we add an extra feature matching loss to encourage the discriminator to get more discriminative information from the reconstructed samples. Our experimental results on the CIC-IDS2018 dataset indicates that compared with autoencoder and generative adversarial network, our model can effectively improve the detection accuracy and can be applied to 6G network traffic security detection. © 2023, ICST Institute for Computer sciences, Social Informatics and Telecommunications Engineering.

关键词： Intrusion detection

来源：评论

学校读者我要写书评

暂无评论

Classification of Deforestation Factors in 6G Satellite Forest Images 5th

Classification of Deforestation Factors in 6G Satellite Fore...

引用

5th EAI International Conference on 6G for Future Wireless Networks, 6GN 2022

作者： Li, Yuhai Sun, Yuxin Meng, Xianglong Xi, Liang Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin300300 China School of Computer Science and Technology Harbin University of Science and Technology Harbin150080 China

ISBN: (纸本)9783031360107

The terrestrial satellite network will play an essential role in 6G. Through the satellite system, people can obtain a lot of ground image information to process tasks and feedback. Forest resource is an essential resource. Determining the causes of deforestation is crucial for the development and implementation of forest protection plans. In this article, we propose a novel deep neural network model for distinguishing drivers of deforestation events from satellite forest images. To solve the problems of image rotation caused by satellite angle rotation and image blurring caused by extreme weather occlusion during satellite image acquisition, we add data enhancement and a self-supervised rotation loss to the model to improve the robustness and adaptability. We use deforestation maps generated from Landsat 8 satellite imagery as a dataset and demonstrate that our approach achieves better results than the baselines. © 2023, ICST Institute for Computer sciences, Social Informatics and Telecommunications Engineering.

关键词： Deforestation

来源：评论

学校读者我要写书评

暂无评论

SDM-based electronic defense intelligent networking technology 8

SDM-based electronic defense intelligent networking technolo...

引用

8th Symposium on Novel Photoelectronic Detection technology and Applications

作者： Ma, Baoqiang Liu, Tongyu Yong, Zhang Shuai, Jia Dong, Liu Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin300308 China

ISBN: (数字)9781510653122

ISBN: (纸本)9781510653115

With the increasingly prominent role of high-tech weapons in modern warfare, electronic warfare has also become a topic that has attracted worldwide attention. At present, global electronic warfare is in an important period of transformation and development. In particular, the US military has made substantial progress in electronic warfare in recent years. The development of electronic defense has played a decisive role in future wars[1]. This article introduces the application system architecture and operating mechanism of SDM. At the same time, it introduces the application of plug-and-play SDM software in modern electronic defense system operations, and expounds the role of intelligent command in future electronic warfare. SDM software can realize the plug-and-play of sensor equipment, and it plays a key role in the intelligent network of electronic defense. © 2022 SPIE

关键词： electronic warfare

来源：评论

学校读者我要写书评

暂无评论

Analysis of pump power distribution in diode pumped Tm:YAP laser 8

Analysis of pump power distribution in diode pumped Tm:YAP l...

引用

8th Symposium on Novel Photoelectronic Detection technology and Applications

作者： Wang, Lixin Xu, Hong Yan, Mengmeng Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin300308 China

ISBN: (数字)9781510653122

ISBN: (纸本)9781510653115

We analyzed the space overlap of laser transverse modes inside cavity of a Tm:YAP laser and derived the formulae of laser pump power distribution on variable harmonic transverse modes. The method for reducing the influences of high-order transverse mode on laser output beam quality are put forward and validated by experiments, the results can be used in the designing of such lasers. © 2022 SPIE

关键词： Pumping (laser)

来源：评论

学校读者我要写书评

暂无评论

Dim and small targets based on the fusion of infrared radiation characteristics and spectral characteristics research on detection methods 8

Dim and small targets based on the fusion of infrared radiat...

引用

8th Symposium on Novel Photoelectronic Detection technology and Applications

作者： Dong, Liu Yang, Yining Liu, Tongyu Yong, Zhang Hu, Yutong Shuai, Jia Ma, Baoqiang Science and Technology on Electro-Optical Information Security Control Laboratory Tianjin300308 China

ISBN: (数字)9781510653122

ISBN: (纸本)9781510653115

The geometric and motion characteristics of false targets are more similar to those of real targets. The detection accuracy of traditional infrared radiation characteristics methods for infrared targets is obviously low. For small infrared weak targets with small size and low signal-to-noise ratio, it is difficult to effectively extract shape features by using a single band detection method. Therefore, this paper makes full use of the advantages of infrared single band image and multispectral image target detection methods, and combines single band detection technology and multispectral detection technology, which not only avoids the defect of single band detection missing alarm, but also reduces the impact of large amount of multispectral detection data, and has real-time performance and good detection effect. The hybrid detection algorithm proposed in this paper has stronger adaptability, higher recognition rate and faster processing ability. © 2022 SPIE

关键词： Infrared radiation

来源：评论

学校读者我要写书评

暂无评论

Study on the Medium Fusion Data Storage technology for High Performance Airborne Sensors 3

Study on the Medium Fusion Data Storage Technology for High ...

引用

3rd IEEE International Conference on Intelligent technology and Embedded Systems, ICITES 2023

作者： Xie, Dan Tian, Haishan Sun, Yuxin Zhang, Mingru Huang, Yawei Tang, Chang School of Physics and Electronics Hunan Normal University Changsha China Science and Technology on Electro-Optical Information Security Control Laboratory Tianjing China

ISBN: (纸本)9798350343755

The data rate of high-performance airborne sensors is much higher than the wireless data transfer rate in many detection applications, so the onboard data storage systems are usually used to store the sensor data. There is a problem in the onboard data storage that the frequency of address jumps among incontinuous areas is very high due to the high data updating frequency of the FAT (File Allocation Table) and FDT (File Directory Table), which would significantly reduce the data storage speed. This is because the most widely used NAND (Not And) flash has a long response time when storing the data in the incontinuous areas. In order to solve the problem, a called medium fusion storage method, which uses multiple mediums time sharing to store the data of file areas and data region, is proposed in this study. The performance of the proposed method is tested on an actual data storage system of the airborne sensor. The experimental results show that medium fusion storage method significantly reduces the data updating frequency of the FAT and FDT in the data storage, and the response time is negligible compared with the actual data writing time. © 2023 IEEE.

关键词： Data transfer

来源：评论

学校读者我要写书评

暂无评论

Design and Implementation of High-Speed Serial Transmission for ADC Based on JESD204B 3

Design and Implementation of High-Speed Serial Transmission ...

引用

3rd IEEE International Conference on Intelligent technology and Embedded Systems, ICITES 2023

作者： Huang, Yawei Tian, Haishan Zhang, Mingru Sun, Yuxin Xie, Dan Tang, Chang School of Physics and Electronics Hunan Normal University Changsha China Science and Technology on Electro-Optical Information Security Control Laboratory Tianjing China

ISBN: (纸本)9798350343755

Currently, as the resolution of sensors such as radar continues to improve, the signal bandwidth is rapidly increasing. There is a growing demand for higher sampling rates and data transmission rates in Analog-to-Digital Converters (ADCs). However, the parallel transmission commonly used in ADCs faces issues such as high crosstalk and difficulties in transmission synchronization. In response to this problem, this paper proposes a method for the design and implementation of a high-speed serial data transmission interface based on JESD204B. Specifically, it utilizes Xilinx's ZYNQ-7 series FPGA and ADC as the core components and presents clock synchronization design, ADC circuit design, and FPGA 204B IP design methods. Experimental testing was conducted using selected link parameters (LMFS=4222) as an example. The experimental results demonstrate that correct JESD204B interface data transmission for four channels was achieved at a data rate of 10 Gbps, thus validating the feasibility of the proposed high-speed transmission link design based on JESD204B in this paper. © 2023 IEEE.

关键词： Field programmable gate arrays (FPGA)

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：