A5078021087- Advanced Image and Video Retrieval Techniques
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Multimodal Machine Learning Applications
- Digital Media Forensic Detection
- Advanced Vision and Imaging
- Domain Adaptation and Few-Shot Learning
- Image Processing Techniques and Applications
- Advanced Neural Network Applications
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Image and Video Quality Assessment
- Visual Attention and Saliency Detection
- Face recognition and analysis
- Gastrointestinal motility and disorders
- Neural Networks and Applications
- Gut microbiota and health
- Reservoir Engineering and Simulation Methods
- Hydrocarbon exploration and reservoir analysis
- Urban Transport and Accessibility
- Geological Modeling and Analysis
- 3D Shape Modeling and Analysis
- Human Pose and Action Recognition
- Computer Graphics and Visualization Techniques
- Machine Learning and ELM
Academy of Broadcasting Science
2018-2024
Communication University of China
2021-2023
Tianjin University of Science and Technology
2019
Xi'an Shiyou University
2019
State Administration of Press, Publication, Radio, Film and Television
2018
Beijing Technology and Business University
2016-2017
Multiple articles have confirmed that an imbalance of the intestinal microbiota is closely related to aberrant immune responses intestines and pathogenesis inflammatory bowel diseases (IBDs).
Video super-resolution is a challenging task, which has attracted great attention in research and industry communities. In this paper, we propose novel end-to-end architecture, called Residual Invertible Spatio-Temporal Network (RISTN) for video super-resolution. The RISTN can sufficiently exploit the spatial information from low-resolution to high-resolution, effectively models temporal consistency consecutive frames. Compared with existing recurrent convolutional network based approaches,...
In media industry, the demand of SDR-to-HDRTV upconversion arises when users possess HDR-WCG (high dynamic range-wide color gamut) TVs while most off-the-shelf footage is still in SDR (standard range). The research community has started tackling this low-level vision task by learning-based approaches. When applied to real SDR, yet, current methods tend produce dim and desaturated result, making nearly no improvement on viewing experience. Different from other network-oriented methods, we...
Abstract Recently, image super‐resolution works based on Convolutional Neural Networks (CNNs) and Generative Adversarial Nets (GANs) have shown promising performance. However, these methods tend to generate blurry over‐smoothed super‐resolved (SR) images, due the incomplete loss function powerless architectures of networks. In this paper, a novel generative adversarial through deep dense skip connections (GSR‐DDNet), is proposed solve above‐mentioned problems. It aims take advantage GAN's...
Media content forgery is widely spread over the Internet and has raised severe societal concerns. With development of deep learning, new technologies such as generative adversarial networks (GANs) media technology have already been utilized for politicians celebrity forgery, which a terrible impact on society. Existing GAN-generated face detection approaches rely detecting image artifacts generated traces. However, these methods are model-specific, performance deteriorated when faced with...
The advancement of deep forgery technology has significantly impacted the credibility media content, making detection forgeries crucial for ensuring security. Although research on deepfake methods been progressively advancing, current approaches predominantly rely detecting and identifying artifacts. As continually improves, high-quality synthetic images those produced through reconstruction have become increasingly sophisticated, rendering artifact trace somewhat limited. To address this...
Deep convolutional neural networks (CNNs) based image super-resolution approaches have reached significant success in recent years. However, due to the information-discarded nature of CNN, they inevitably suffer from information loss during feature embedding process, which extracted intermediate features cannot effectively represent or reconstruct input. As a result, super-resolved will large deviations structure with its low-resolution version, leading inaccurate representations some local...
Nowadays, deepfake detection on subtle-expression manipulation, facial-detail modification, and smeared images has become a research hotspot. Existing deepfake-detection methods the whole face are coarse-grained, where details missing due to negligible manipulated size of image. To address problems, we propose build transformer model for method by organ, obtain features. We reduce weight defaced or unclear organs prioritize clear intact organs. Meanwhile, simulate real-world environment,...
Sketch-based 3D shape retrieval has unique representation availability of the queries and vast applications. Therefore, it received more attentions in research community content-based object retrieval. However, sketch-based is a challenging topic due to semantic gap existing between inaccurate sketches accurate models. In order enrich advance study retrieval, we initialize on model collect sketch dataset based developed sketching interface which facilitates us draw air while standing front...
Image manipulation methods, such as the copy-move, splicing, and removal have become increasingly mature changed common perception of “seeing is believing.” The credibility digital media has been seriously damaged with development image methods. Most detection methods detect traces tampering pixel by pixel. As a result, detected areas are separated, which results in insufficient consideration content at object level. In this paper, based on forgery discrimination proposed. Specifically,...
Existing image captioning methods are always limited to the rules of words or syntax with single sentence and poor words. In this paper, paper introduces a novel framework for tasks which reconciles slot filling approaches neural network approaches. Our approach first generates template many locations using Wasserstein Generative Adversarial Network (WGAN). Then slots in visual regions will be filled by object detectors. model consists structured generator multi-level discriminator....
Recent convolutional neural network (CNNs) have shown promising performance on image retrieval due to the powerful feature extraction capability. However, potential relations of maps are not effectively exploited in before CNNs, resulting inaccurate representations. To address this issue, we excavate channel-wise realtions by a matching strategy adaptively highlight informative features. In paper, propose novel representative (RFMN) for hashing retrieval. Specifically, block (RFMB) that can...
In media industry, the demand of SDR-to-HDRTV up-conversion arises when users possess HDR-WCG (high dynamic range-wide color gamut) TVs while most off-the-shelf footage is still in SDR (standard range). The research community has started tackling this low-level vision task by learning-based approaches. When applied to real SDR, yet, current methods tend produce dim and desaturated result, making nearly no improvement on viewing experience. Different from other network-oriented methods, we...
At present, the development of deep forgery technology has brought new challenges to media content forensics, and use identification methods identify forged audio video become a significant focus research difficulty. Deep forensic play mutual game promote each other’s development. This paper proposes spatiotemporal local feature abstraction (STLFA) framework for facial solve industry technology. To adequately utilize features, we combine key points, point movement, corner points detect...
Image retrieval based on deep convolutional neural networks (CNNs) has achieved promising performance in recent years. However, there are many low-resolution images real-world tasks, and they would result inaccurate hash representations for the CNN, which only trained high-resolution images. In this paper, we propose a novel framework, is called superresolution hashing (DSR-Hashing), to solve problem. DSR-Hashing constructed by two components: super-resolution network an encoding network....