Yuming Zhao

ORCID: 0009-0007-1443-3463
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Image and Video Retrieval Techniques
  • Advanced Vision and Imaging
  • Image Retrieval and Classification Techniques
  • Topic Modeling
  • Video Surveillance and Tracking Methods
  • Remote Sensing and Land Use
  • Advanced Neural Network Applications
  • Medical Image Segmentation Techniques
  • Natural Language Processing Techniques
  • Face and Expression Recognition
  • Computer Graphics and Visualization Techniques
  • Image Enhancement Techniques
  • Face recognition and analysis
  • Video Analysis and Summarization
  • Human Pose and Action Recognition
  • Generative Adversarial Networks and Image Synthesis
  • Domain Adaptation and Few-Shot Learning
  • Image and Video Stabilization
  • Advanced Algorithms and Applications
  • Image Processing Techniques and Applications
  • Data Quality and Management
  • QR Code Applications and Technologies
  • Multimodal Machine Learning Applications
  • Image and Object Detection Techniques
  • Advanced Measurement and Detection Methods

Northwestern Polytechnical University
2025

Shanghai Jiao Tong University
2011-2024

Shenyang Institute of Automation
2024

Chinese Academy of Sciences
2024

State Key Laboratory of Robotics
2024

Jingdong (China)
2022-2023

Detection Limit (United States)
2020

Shanxi Agricultural University
2007

10.1016/j.jvcir.2009.07.006 article EN Journal of Visual Communication and Image Representation 2009-07-29

The legal charge prediction task aims to judge appropriate charges according the given fact description in cases. Most existing methods formulate it as a multi-class text classification problem and have achieved tremendous progress. However, performance on low-frequency is still unsatisfactory. Previous studies indicate leveraging label information can facilitate this task, but approaches utilizing are not fully explored. In paper, inspired by vision-language fusion techniques multi-modal...

10.1145/3511808.3557379 article EN Proceedings of the 31st ACM International Conference on Information & Knowledge Management 2022-10-16

This paper presents an efficient algorithm to track the object contour in image sequences. Firstly, probability distributions of color feature and texture are approximated by different schemes. Secondly, background pixels distinguished calculating their posterior light Bayesian formula based on two models. Finally, energy functional is proposed solved gradient descent flow evolve initial which estimated motion feature, so that minimum corresponds contour. In order save computational...

10.1142/s0218001409007600 article EN International Journal of Pattern Recognition and Artificial Intelligence 2009-11-01

This paper presents a three-step framework to remove the highlight exists on objects in certain conditions. Unlike traditional HDR (High Dynamic Range) technology requires multiple registrated image sequence; our method needs only two arbitrary images. SURF (Speeded Up Robust Features) matching algorithm is first applied find corresponding point pairs between images; homography then found by perspective transformation theory; minimum gray selection used at last eliminated and fuse related...

10.1109/icmult.2010.5631281 article EN 2010-10-01

Dialogue representation and understanding aim to convert conversational inputs into embeddings fulfill discriminative tasks. Compared with free-form text, dialogue has two important characteristics, hierarchical semantic structure multi-facet attributes. Therefore, directly applying the pretrained language models (PLMs) might result in unsatisfactory performance. Recently, several work focused on dialogue-adaptive post-training (DialPost) that further trains PLMs fit dialogues. To model...

10.18653/v1/2023.acl-long.564 article EN cc-by 2023-01-01

As an important and fundamental methodology in the fields of pattern recognition image processing, learning middle level feature has attracted increasing interest during recent years, where generative mapping shown highly completive performance diverse applications. In this paper, a representation is proposed based on Deep Boltzmann Machine (DBM) sufficient statistics (SS) for detection. approach, DBM employed to model data distribution hidden information inferred by together with other...

10.1109/icpr.2014.152 article EN 2014-08-01

Pedestrian detection and recognition has become the basic research in various social fields. Convolutional neural networks have excellent learning ability can recognize patterns with robustness to some extent distortions transformations. Yet, they need much more intermediate hidden units cannot from unlabeled samples. In this paper, we purpose a latent training model based on convolutional network. The purposed adopts part detectors reduce scale of layer. It also follows method determine...

10.1109/icedif.2015.7280162 article EN 2015-01-01

Texture information plays an important role in rendering true objects, especially with the wide application of image-based three-dimensional (3-D) reconstruction and 3-D laser scanning. This paper proposes a seamless texture mapping algorithm to achieve high-quality visual effect for reconstruction. At first, series image sets is produced by analyzing visibility triangular facets, are clustered segmented into number optimal reference patches. Second, generated patches sequenced create rough...

10.1117/1.jei.25.5.053025 article EN Journal of Electronic Imaging 2016-10-11

In this paper we propose a novel multiple target tracking model composed of two detectors and tracker. An on-line detector tracker are used to generate candidates, whose confidence scores then evaluated by the off-line trained detectors. data association stage, high-efficient inference in structural leads optimal result. The experimental results demonstrate that our can overcome occlusion appearance changing problems. be applied analyze information single or among targets.

10.1109/icme.2014.6890133 article EN 2022 IEEE International Conference on Multimedia and Expo (ICME) 2014-07-01

Recent cross-lingual summarization research has pursued the use of a unified end-to-end model which demonstrated certain level improvement in performance and effectiveness, but this approach stitches together multiple tasks makes computation more complex. Less work focused on alignment relationships across languages, led to persistent problems summary misordering loss key information. For reason, we first simplify multitasking by converting translation task into an equal proportion so that...

10.3390/app13116723 article EN cc-by Applied Sciences 2023-05-31

In scene parsing, the model is required to be able process complex multi-modal data such as images and contexts in real scenes, discover their implicit connections from objects existing scene. As a storage method that contains entity information relationship between entities, knowledge graph can well express semantic this paper, new multi-phase was proposed solve parsing tasks; first, used align then graph-based generates results. We also designed an experiment of feature engineering’s...

10.3390/app13127115 article EN cc-by Applied Sciences 2023-06-14

With its unique information-filtering function, text summarization technology has become a significant aspect of search engines and question-and-answer systems. However, existing models that include the copy mechanism often lack ability to extract important fragments, resulting in generated content suffers from thematic deviation insufficient generalization. Specifically, Chinese automatic using traditional generation methods loses semantics because reliance on word lists. To address these...

10.1145/3643695 article EN ACM Transactions on Asian and Low-Resource Language Information Processing 2024-02-06
Coming Soon ...