Xianhui Lin

ORCID: 0000-0002-8974-2064
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Image Processing Techniques
  • Generative Adversarial Networks and Image Synthesis
  • Human Motion and Animation
  • Face recognition and analysis
  • Image and Video Quality Assessment
  • Visual Attention and Saliency Detection
  • Image Enhancement Techniques
  • Advanced Vision and Imaging
  • Video Analysis and Summarization
  • Multimedia Communication and Technology
  • Image Processing Techniques and Applications
  • Image and Signal Denoising Methods
  • Subtitles and Audiovisual Media
  • Advanced Neural Network Applications
  • MRI in cancer diagnosis
  • Endometriosis Research and Treatment
  • Robotics and Sensor-Based Localization
  • Digital Humanities and Scholarship
  • Digital Media and Visual Art
  • Uterine Myomas and Treatments
  • Galectins and Cancer Biology
  • COVID-19 and healthcare impacts
  • Robotic Mechanisms and Dynamics
  • Gynecological conditions and treatments
  • Autoimmune and Inflammatory Disorders Research

Alibaba Group (China)
2020-2024

Second Affiliated Hospital & Yuying Children's Hospital of Wenzhou Medical University
2022-2024

Wenzhou Medical University
2022-2024

Alibaba Group (Cayman Islands)
2021-2023

Nanjing University of Science and Technology
2020

Alibaba Group (United States)
2020

South China University of Technology
2017

Wuhan University of Technology
2013

Face restoration is important in face image processing, and has been widely studied recent years. However, previous works often fail to generate plausible high quality (HQ) results for real-world low (LQ) images. In this paper, we propose a new progressive semantic-aware style transformation framework, named PSFR-GAN, restoration. Specifically, instead of using an encoder-decoder framework as methods, formulate the LQ images multi-scale procedure through transformation. Given pair its...

10.1109/cvpr46437.2021.01172 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2021-06-01

This paper reviews the video colorization challenge on New Trends in Image Restoration and Enhancement (NTIRE) workshop, held conjunction with CVPR 2023. The target of this is converting grayscale videos into color better performance temporal consistency. consists two tracks. For Track 1, goal achieving best FID (Fréchet Inception Distance) while being constrained to maintain or improve over baseline method terms temporal-consistency metric. Color Distribution Consistency (CDC) index used as...

10.1109/cvprw59228.2023.00159 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2023-06-01

Text-to-image diffusion models (T2I) have demonstrated unprecedented capabilities in creating realistic and aesthetic images. On the contrary, text-to-video (T2V) still lag far behind frame quality text alignment, owing to insufficient quantity of training videos. In this paper, we introduce VideoElevator, a training-free plug-and-play method, which elevates performance T2V using superior T2I. Different from conventional sampling (i.e., temporal spatial modeling), VideoElevator explicitly...

10.1609/aaai.v39i10.33114 article EN Proceedings of the AAAI Conference on Artificial Intelligence 2025-04-11

Few-shot font generation is challenging, as it needs to capture the fine-grained stroke styles from a limited set of reference glyphs, and then transfer other characters, which are expected have similar styles. However, due diversity complexity Chinese styles, synthesized glyphs existing methods usually exhibit visible artifacts, such missing details distorted strokes. In this paper, we propose VQGAN-based framework (i.e., VQ-Font) enhance glyph fidelity through token prior refinement...

10.1609/aaai.v38i15.29577 article EN Proceedings of the AAAI Conference on Artificial Intelligence 2024-03-24

Jun-Yan He, Zhi-Qi Cheng, Chenyang Li, Jingdong Sun, Wangmeng Xiang, Xianhui Lin, Xiaoyang Kang, Zengke Jin, Yusen Hu, Bin Luo, Yifeng Geng, Xuansong Xie. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track. 2023.

10.18653/v1/2023.emnlp-industry.23 article EN cc-by 2023-01-01

This paper introduces the WordArt Designer API, a novel framework for user-driven artistic typography synthesis utilizing Large Language Models (LLMs) on ModelScope. We address challenge of simplifying non-professionals by offering dynamic, adaptive, and computationally efficient alternative to traditional rigid templates. Our approach leverages power LLMs understand interpret user input, facilitating more intuitive design process. demonstrate through various case studies how users can...

10.48550/arxiv.2401.01699 preprint EN other-oa arXiv (Cornell University) 2024-01-01

Objectives This study assessed the clinical value of parameters derived from dynamic contrast-enhanced (DCE) MRI with respect to correlation angiogenesis and proliferation cervical cancer, performance diagnosis reproducibility DCE-MRI across scanners. Materials Methods A total 113 patients carcinoma two centers were included in this retrospective study. The DCE data centralized processed using five tracer kinetic models (TKMs) (Tofts, Ex-Tofts, ATH, SC, DP), yielding following parameters:...

10.3389/fonc.2022.958219 article EN cc-by Frontiers in Oncology 2022-10-17

Gastric cancer (GC) has high rates of morbidity and mortality, this phenomenon is particularly evident in coastal regions where local dietary habits favor the consumption pickled foods such as salted fish vegetables. In addition, diagnosis rate GC remains low due to lack diagnostic serum biomarkers. Therefore, study, we aimed identify potential biomarkers for use clinical practice. To candidate GC, 88 samples were first screened using a high-throughput protein microarray measure levels 640...

10.1111/cas.15876 article EN cc-by-nc Cancer Science 2023-06-08

Text-to-image diffusion models (T2I) have demonstrated unprecedented capabilities in creating realistic and aesthetic images. On the contrary, text-to-video (T2V) still lag far behind frame quality text alignment, owing to insufficient quantity of training videos. In this paper, we introduce VideoElevator, a training-free plug-and-play method, which elevates performance T2V using superior T2I. Different from conventional sampling (i.e., temporal spatial modeling), VideoElevator explicitly...

10.48550/arxiv.2403.05438 preprint EN arXiv (Cornell University) 2024-03-08

Human visual imagination usually begins with analogies or rough sketches. For example, given an image a girl playing guitar before building, one may analogously imagine how it seems like if Iron Man Pyramid in Egypt. Nonetheless, condition not be precisely aligned the imaginary result indicated by text prompt, and existing layout-controllable text-to-image (T2I) generation models is prone to producing degraded generated results obvious artifacts. To address this issue, we present novel T2I...

10.48550/arxiv.2404.06451 preprint EN arXiv (Cornell University) 2024-04-09

MetaDesigner revolutionizes artistic typography synthesis by leveraging the strengths of Large Language Models (LLMs) to drive a design paradigm centered around user engagement. At core this framework lies multi-agent system comprising Pipeline, Glyph, and Texture agents, which collectively enable creation customized WordArt, ranging from semantic enhancements imposition complex textures. incorporates comprehensive feedback mechanism that harnesses insights multimodal models evaluations...

10.48550/arxiv.2406.19859 preprint EN arXiv (Cornell University) 2024-06-28

Background Diffuse uterine leiomyomatosis (DUL) is a seldom-seen condition, with only handful of cases magnetic resonance imaging (MRI) findings documented. In clinical settings, it often mistaken for multiple leiomyomas due to lack adequate recognition DUL. Objective This study shows two instances DUL, underscoring their MRI improve preoperative diagnostic precision. Conclusion For patients exhibiting masses present in the parametrial and abdominal cavities, consideration should be given...

10.3389/fonc.2024.1430531 article EN cc-by Frontiers in Oncology 2024-07-03

Recent reference-based face restoration methods have received considerable attention due to their great capability in recovering high-frequency details on real low-quality images. However, most of these require a high-quality reference image the same identity, making them only applicable limited scenes. To address this issue, paper suggests deep dictionary network (termed as DFDNet) guide process degraded observations. begin with, we use K-means generate dictionaries for perceptually...

10.48550/arxiv.2008.00418 preprint EN other-oa arXiv (Cornell University) 2020-01-01

Face restoration is important in face image processing, and has been widely studied recent years. However, previous works often fail to generate plausible high quality (HQ) results for real-world low (LQ) images. In this paper, we propose a new progressive semantic-aware style transformation framework, named PSFR-GAN, restoration. Specifically, instead of using an encoder-decoder framework as methods, formulate the LQ images multi-scale procedure through transformation. Given pair its...

10.48550/arxiv.2009.08709 preprint EN cc-by-nc-sa arXiv (Cornell University) 2020-01-01

Few-shot font generation is challenging, as it needs to capture the fine-grained stroke styles from a limited set of reference glyphs, and then transfer other characters, which are expected have similar styles. However, due diversity complexity Chinese styles, synthesized glyphs existing methods usually exhibit visible artifacts, such missing details distorted strokes. In this paper, we propose VQGAN-based framework (i.e., VQ-Font) enhance glyph fidelity through token prior refinement...

10.48550/arxiv.2308.14018 preprint EN cc-by arXiv (Cornell University) 2023-01-01

The problem of robust alignment batches images can be formulated as a low-rank matrix optimization problem, relying on the similarity well-aligned images. Going further, observing that to aligned are sampled from union subspaces, we propose new method based subspace recovery techniques provide more and accurate alignment. proposed seeks set domain transformations which applied unaligned so resulting made similar possible. linearized series convex problems solved by alternative sparsity...

10.1007/s41095-017-0080-x article EN cc-by Computational Visual Media 2017-05-05

Cloud manufacturing employs some progressive issues, for example, the ideas of cloud computing. It extends "Software as a Service" concept to "Manufacture Service", sharing "manufacturing resources and capability". Access control prevents illegal operations effectively by monitoring user's behavior. This paper proposes an access model environment that called WRBAC. Considering architectures characteristics manufacturing, uses layered structure achieve system-level control. Based on ABAC,...

10.1109/dasc.2013.39 article EN 2013-12-01

Most common publications are related to COVID-19 diagnosis in hematological malignancy patients. However, here we report a case involving patient diagnosed with B-cell lymphoma while undergoing treatment for COVID-19, including the changes major clinical symptoms and medical examinations, then explain probable causes of case.A 74-year-old woman previous history oesophageal cancer was admitted hospital after having cough sputum 15 days. Despite symptoms, this did not have fever at time onset....

10.2174/1573405618666220329210311 article EN Current Medical Imaging Formerly Current Medical Imaging Reviews 2022-03-30

This paper introduces WordArt Designer, a user-driven framework for artistic typography synthesis, relying on the Large Language Model (LLM). The system incorporates four key modules: LLM Engine, SemTypo, StyTypo, and TexTypo modules. 1) empowered by (e.g., GPT-3.5), interprets user inputs generates actionable prompts other modules, thereby transforming abstract concepts into tangible designs. 2) SemTypo module optimizes font designs using semantic concepts, striking balance between...

10.48550/arxiv.2310.18332 preprint EN other-oa arXiv (Cornell University) 2023-01-01

In this paper, we present DreaMoving, a diffusion-based controllable video generation framework to produce high-quality customized human videos. Specifically, given target identity and posture sequences, DreaMoving can generate of the moving or dancing anywhere driven by sequences. To end, propose Video ControlNet for motion-controlling Content Guider preserving. The proposed model is easy use be adapted most stylized diffusion models diverse results. project page available at...

10.48550/arxiv.2312.05107 preprint EN other-oa arXiv (Cornell University) 2023-01-01
Coming Soon ...