作者：
王振华吴伟仁田玉龙田金文柳健Institute for Pattern Recognition and Artificial Intelligence
State Key Lab for Image Processing and Intelligent ControlHuazhong University of Science and Technology Wuhan 430074 China Institute for Pattern Recognition and Artificial Intelligence
State Key Lab for Image Processing and Intelligent ControlHuazhong University of Science and Technology Wuhan 430074 China major limitation for deep space communication is the limited bandwidths available. The downlink rate using X-band with an L2 halo orbit is estimated to be of only 5.35 GB/d. However the Next Generation Space Telescope (NGST) will produce about 600 GB/d. Clearly the volume of data to downlink must be reduced by at least a factor of 100. One of the resolutions is to encode the data using very low bit rate image compression techniques. An very low bit rate image compression method based on region of interest(ROI) has been proposed for deep space image. The conventional image compression algorithms which encode the original data without any data analysis can maintain very good details and haven't high compression rate while the modern image compressions with semantic organization can have high compression rate even to be hundred and can't maintain too much details. The algorithms based on region of interest inheriting from the two previews algorithms have good semantic features and high fidelity and is therefore suitable for applications at a low bit rate. The proposed method extracts the region of interest by texture analysis after wavelet transform and gains optimal local quality with bit rate control. The Result shows that our method can maintain more details in ROI than general image compression algorithm(SPIHT) under the condition of sacrificing the quality of other uninterested areas
A major limitation for deep space communication is the limited bandwidths available. The downlinkrate using X-band with an L2 halo orbit is estimated to be of only 5.35 GB/d. However, the Next GenerationSpace Telescop...
详细信息
A major limitation for deep space communication is the limited bandwidths available. The downlinkrate using X-band with an L2 halo orbit is estimated to be of only 5.35 GB/d. However, the Next GenerationSpace Telescope (NGST) will produce about 600 GB/d. Clearly the volume of data to downlink must be re-duced by at least a factor of 100. One of the resolutions is to encode the data using very low bit rate image com-pression techniques. An very low bit rate image compression method based on region of interest(ROI) has beenproposed for deep space image. The conventional image compression algorithms which encode the original datawithout any data analysis can maintain very good details and haven' t high compression rate while the modernimage compressions with semantic organization can have high compression rate even to be hundred and can' tmaintain too much details. The algorithms based on region of interest inheriting from the two previews algorithmshave good semantic features and high fidelity, and is therefore suitable for applications at a low bit rate. Theproposed method extracts the region of interest by texture analysis after wavelet transform and gains optimal localquality with bit rate control. The Result shows that our method can maintain more details in ROI than generalimage compression algorithm(SPIHT) under the condition of sacrificing the quality of other uninterested areas.
Named Entity Recognition (NER) is an important task in knowledge extraction, which targets extracting structural information from unstructured text. To fully employ the prior-knowledge of the pre-trained language mode...
详细信息
Named Entity Recognition (NER) is an important task in knowledge extraction, which targets extracting structural information from unstructured text. To fully employ the prior-knowledge of the pre-trained language models, some research works formulate the NER task into the machine reading comprehension form (MRC-form) to enhance their model generalization capability of commonsense knowledge. However, this transformation still faces the data-hungry issue with limited training data for the specific NER tasks. To address the low-resource issue in NER, we introduce a method named active multi-task-based NER (AMT-NER), which is a two-stage multi-task active learning training model. Specifically, A multi-task learning module is first introduced into AMT-NER to improve its representation capability in low-resource NER tasks. Then, a two-stage training strategy is proposed to optimize AMT-NER multi-task learning. An associated task of Natural Language Inference (NLI) is also employed to enhance its commonsense knowledge further. More importantly, AMT-NER introduces an active learning module, uncertainty selective, to actively filter training data to help the NER model learn efficiently. Besides, we also find different external supportive data under different pipelines improves model performance differently in the NER tasks. Extensive experiments are performed to show the superiority of our method, which also proves our findings that the introduction of external knowledge is significant and effective in the MRC-form NER tasks.
With the emergence of edge computing, there’s a growing need for advanced technologies capable of real-time, efficient processing of complex data on edge devices, particularly in mobile health systems handling pathol...
详细信息
With the emergence of edge computing, there’s a growing need for advanced technologies capable of real-time, efficient processing of complex data on edge devices, particularly in mobile health systems handling pathological images. On edge computing devices, the lightweighting of models and reduction of computational requirements not only save resources but also increase inference speed. Although many lightweight models and methods have been proposed in recent years, they still face many common challenges. This paper introduces a novel convolution operation, Dynamic Scalable Convolution (DSC), which optimizes computational resources and accelerates inference on edge computing devices. DSC is shown to outperform traditional convolution methods in terms of parameter efficiency, computational speed, and overall performance, through comparative analyses in computer vision tasks like image classification and semantic segmentation. Experimental results demonstrate the significant potential of DSC in enhancing deep neural networks, particularly for edge computing applications in smart devices and remote healthcare, where it addresses the challenge of limited resources by reducing computational demands and improving inference speed. By integrating advanced convolution technology and edge computing applications, DSC offers a promising approach to support the rapidly developing mobile health field, especially in enhancing remote healthcare delivery through mobile multimedia communication.
The rapid development of computer vision technology for detecting anomalies in industrial products has received unprecedented attention. In this paper, we propose a dual teacher–student-based discrimination model (DT...
详细信息
The rapid development of computer vision technology for detecting anomalies in industrial products has received unprecedented attention. In this paper, we propose a dual teacher–student-based discrimination model (DTSD) for anomaly detection, which combines the advantages of both embedding-based and reconstruction-based methods. First, the DTSD builds a dual teacher-student architecture consisting of a pretrained teacher encoder with frozen parameters, a student encoder and a student decoder. By distillation of knowledge from the teacher encoder, the two teacher-student modules acquire the ability to capture both local and global anomaly patterns. Second, to address the issue of poor reconstruction quality faced by previous reconstruction-based approaches in some challenging cases, the model employs a feature bank that stores encoded features of normal samples. By incorporating template features from the feature bank, the student decoder receives explicit guidance to enhance the quality of reconstruction. Finally, a segmentation network is utilized to adaptively integrate multiscale anomaly information from the two teacher–student modules, thereby improving segmentation accuracy. Extensive experiments demonstrate that our method outperforms existing state-of-the-art approaches. The code of DTSD is publicly available on https://***/Math-Computer/DTSD.
暂无评论