A5110382448- Video Coding and Compression Technologies
- Advanced Data Compression Techniques
- Advanced Vision and Imaging
- Topic Modeling
- Advanced Image Processing Techniques
- Wireless Communication Security Techniques
- Natural Language Processing Techniques
- Chaos-based Image/Signal Encryption
- Image and Video Quality Assessment
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Digital Media Forensic Detection
- Error Correcting Code Techniques
- Advanced Steganography and Watermarking Techniques
- IoT and Edge/Fog Computing
- Video Analysis and Summarization
- Algorithms and Data Compression
- Sparse and Compressive Sensing Techniques
- Advanced Text Analysis Techniques
- Cooperative Communication and Network Coding
- Parallel Computing and Optimization Techniques
- Image and Object Detection Techniques
- Digital Filter Design and Implementation
- Cellular Automata and Applications
- Cloud Data Security Solutions
Children's Hospital of Chongqing Medical University
2016-2025
First Hospital of Jilin University
2025
Jilin University
2025
Beijing Jiaotong University
2012-2024
Guangdong University of Technology
2024
Nanchang University
2022
University of Jinan
2022
Google (United States)
2020-2021
Henan University of Science and Technology
2021
Brain (Germany)
2021
Recent work pre-training Transformers with self-supervised objectives on large text corpora has shown great success when fine-tuned downstream NLP tasks including summarization. However, tailored for abstractive summarization have not been explored. Furthermore there is a lack of systematic evaluation across diverse domains. In this work, we propose Transformer-based encoder-decoder models massive new objective. PEGASUS, important sentences are removed/masked from an input document and...
Question generation, the task of automatically creating questions that can be answered by a certain span text within given passage, is important for question-answering and conversational systems in digital assistants such as Alexa, Cortana, Google Assistant Siri. Recent sequence to neural models have outperformed previous rule-based systems. Existing mainly focused on using one or two sentences input. Long has posed challenges question generation – worse performances were reported if whole...
A growing number of multi-view data arises naturally in many scenarios, including medical diagnosis, webpage classification, and multimedia analysis. challenge learning from is that not all instances are fully represented views, resulting missing view data. In this paper, we focus on feature-level completion for Aiming at capturing both semantic complementarity identical distribution among different an Isomorphic Linear Correlation Analysis (ILCA) method proposed to linearly map a...
Independent component analysis with soft reconstruction cost (RICA) has been recently proposed to linearly learn sparse representation an overcomplete basis, and this technique exhibits promising performance even on unwhitened data. However, linear RICA may not be effective for the majority of real-world data because nonlinearly separable structure pervasively exists in original space. Meanwhile, is essentially unsupervised method does employ class information. Motivated by success kernel...
In this paper, we present an experimental comparison of 3D rotation techniques. all techniques, the third degree freedom is controlled by mouse-wheel. The investigated techniques are Bell's Trackball, Shoemake's Arcball and Two-axis Valuator method. result from a pilot showed no performance or accuracy difference among these three. However, did observe minor differences in experiment with more participants trials, though were not significant. Also, questionnaires, found that most users...
The tropical Asian and Australasian floras have a close relationship, is vital distribution pattern of seed plants worldwide. As estimated, more than 81 families 225 genera distributed between Asia Australasia. However, the evolutionary dynamics two were still vague. Here, total 29 plant lineages, represented main clades different habits, selected to investigate biotic interchange Australasia by integrated dated phylogenies, biogeography, ancestral state reconstructions. Our statistics...
Interframe and interview predictions are widely employed in multiview video coding. This technique improves the coding efficiency, but it also increases vulnerability of coded bitstream. Thus, one packet loss will affect many subsequent frames same view probably other referenced views. To address this problem, a region-based multiple description scheme is proposed for robust 3-D communication paper, which two descriptions formed by setting left right as dominant first second description,...
For real-time video streaming systems, the packets arriving after display deadline of their frames are considered as late-arrival packets, and typically they discarded. This will affect current frame following ones due to error propagations. this reason, in paper, we propose an approach exploit out-of-order which includes two mechanisms. The first mechanism use these update reference make them more consistent with encoder side, eventually reduce second increase chance successfully decoding...
In this paper, we propose a depth map down-sampling and coding scheme that minimizes the view synthesis distortion. Moreover, solution for optimal problem depth-caused distortion in virtual by exploiting associated texture information along with up-sampling method to be used decoder side is derived. Furthermore, enhance compression performance, synthesized distortion, which evaluated emulating interpolation process, optimization objective function mode selection video encoder. Experimental...
Recently, region-based 3D video coding has been proposed. However, existing view synthesis distortion estimation (VSDE) methods are performed at the frame level. To guide rate-distortion optimization process of schemes, this paper proposes first pixel-level VSDE (PL-VSDE) method. We give definition distortion. estimate it, a backward prediction method is then developed, which starts from pixels interest (POIs) in virtual and finds their corresponding reference via coarse-to-fine approach,...
Abstract This paper proposes a scheme for error-resilient transmission of videos which jointly uses intra macroblock refreshment and redundant motion vector. The selection using or vector is determined by the rate-distortion optimization procedure. end-to-end distortion used optimization, can be easily calculated with recursive optimal per-pixel estimate (ROPE) method. Simulation results show that proposed method outperforms both approach significantly, when two approaches are deployed...
A novel standard-compliant multiple description (MD) video codec is proposed in this paper, which aims to achieve effective redundancy allocation using inter- and intra-description correlation. The inter-description correlation at macro block (MB) level applied produce side information of different modes helpful for better decoding quality. Furthermore, the MB exploited design adaptive skip mode higher compression efficiency. experimental results exhibit a rate central distortion performance...