Yan Yan

ORCID: 0000-0002-3182-1739
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Image and Video Retrieval Techniques
  • Face and Expression Recognition
  • Image Retrieval and Classification Techniques
  • Video Analysis and Summarization
  • Video Surveillance and Tracking Methods
  • Robotics and Sensor-Based Localization
  • Domain Adaptation and Few-Shot Learning
  • Sparse and Compressive Sensing Techniques
  • Advanced Vision and Imaging
  • Aesthetic Perception and Analysis
  • Indoor and Outdoor Localization Technologies
  • Multimodal Machine Learning Applications
  • Human Pose and Action Recognition
  • Face recognition and analysis
  • Advanced Computing and Algorithms
  • Visual Attention and Saliency Detection
  • Color perception and design
  • Topic Modeling
  • Neural dynamics and brain function
  • Blind Source Separation Techniques
  • Color Science and Applications
  • Generative Adversarial Networks and Image Synthesis
  • Computational Geometry and Mesh Generation
  • Machine Learning and Algorithms
  • Advanced Image Processing Techniques

Queen's University
2022

University of Trento
2012-2019

Advanced Digital Sciences Center
2016

Shanghai Jiao Tong University
2013

Shaanxi Normal University
2012

Beihang University
2011

Carnegie Mellon University
2003

People go to fortune tellers in hopes of learning things about their future. A future career path is one the topics most frequently discussed. But rather than rely on "black arts" make predictions, this work we scientifically and systematically study feasibility prediction from social network data. In particular, seamlessly fuse information multiple networks comprehensively describe a user characterize progressive properties his or her path. This accomplished via multi-source framework with...

10.1609/aaai.v30i1.9969 article EN Proceedings of the AAAI Conference on Artificial Intelligence 2016-02-21

Modeling the aging process of human face is important for cross-age verification and recognition. In this paper, we introduce a recurrent (RFA) framework based on neural network which can identify ages people from 0 to 80. Due lack labeled data same person captured in long range ages, traditional models usually split into discrete groups learn one-step feature transformation each pair adjacent age groups. However, those methods neglect in-between evolving states between synthesized faces...

10.1109/cvpr.2016.261 article EN 2016-06-01

In multimedia annotation, due to the time constraints and tediousness of manual tagging, it is quite common utilize both tagged untagged data improve performance supervised learning when only limited training are available. This often done by adding a geometry-based regularization term in objective function model. this case, similarity graph indispensable exploit geometrical relationships among points, construction scheme essentially determines these graph-based algorithms. However, most...

10.1109/tip.2016.2601260 article EN IEEE Transactions on Image Processing 2016-08-18

Labeling video data is an essential prerequisite for many vision applications that depend on training data, such as visual information retrieval, object recognition, and human activity modelling. However, manually creating labels not only time-consuming but also subject to errors, eventually, becomes impossible a very large amount of (e.g. 24/7 surveillance video). To minimize the effort in labeling, we propose unified multiclass active learning approach automatically labeling data. We...

10.1109/iccv.2003.1238391 article EN 2003-01-01

In multimedia annotation, due to the time constraints and tediousness of manual tagging, it is quite common utilize both tagged untagged data improve performance supervised learning when only limited training are available. This often done by adding a geometrically based regularization term in objective function model. this case, similarity graph indispensable exploit geometrical relationships among points, construction scheme essentially determines these graph-based algorithms. However,...

10.1109/cvpr.2015.7299066 article EN 2015-06-01

Multiview action recognition has received increasing attention over the past decade. Various approaches have been proposed to extract view-invariant features; among them, self-similarity matrices (SSMs) shown outstanding performance. However, SSMs become sensitive when there's a very large view change. To make more robust viewpoint changes, authors propose collaborative sparse coding framework. They integrate classifier training process and into unified filtering framework; this lets...

10.1109/mmul.2016.69 article EN IEEE Multimedia 2016-10-01

Correlates between social attention and personality traits have been widely acknowledged in psychology studies. Head pose has commonly employed as a proxy for determining the direction small group interactions. However, impact of head estimation errors on estimates not studied to our knowledge.

10.1145/2522848.2522862 article EN 2013-11-27

Attributes, as mid-level features, have demonstrated great potential in visual recognition tasks due to their excellent propagation capability through different categories. However, existing attribute learning methods are prone the correlated attributes. To discover genuine specific many feature selection been proposed. these implemented at level of raw features that might be very noisy, and usually fail consider structural information space. address this issue, paper, we propose a label...

10.1109/tip.2016.2523340 article EN IEEE Transactions on Image Processing 2016-01-28

10.1016/j.jvcir.2017.02.011 article EN Journal of Visual Communication and Image Representation 2017-02-21

Color plays an essential role in everyday life and is one of the most important visual cues human perception. In abstract art, color means to convey artist's intention affect viewer emotionally. However, colors are rarely experienced isolation, rather, they usually presented together with other colors. fact, expressive properties two-color combinations have been extensively studied by artists. It intriguing try understand how paintings might emotionally, investigate if a computer algorithm...

10.1145/2733373.2806250 article EN 2015-10-13

There is an increasing interest in using hash codes for efficient multimedia retrieval and data storage. The functions are learned such a way that the can preserve essential properties of original space or label information. Then Hamming distance approximate similarity. Existing works have demonstrated success many supervised hashing models. However, labeling time labor consuming, especially scalable datasets. In order to utilize models improve discriminative power codes, we propose...

10.1145/2733373.2806341 article EN 2015-10-13

Community-based health services have risen as important online resources for resolving users concerns. Despite the value, gap between what seekers with specific needs and busy physicians attitudes expertise can offer is being widened. To bridge this gap, we present a question routing scheme that able to connect right physicians. In scheme, first matching via probabilistic fusion of physician-expertise distribution expertise-question distribution. The distributions are calculated by...

10.1109/tcyb.2016.2577590 article EN IEEE Transactions on Cybernetics 2016-06-23

Image-based localization is an essential complement to GPS localization. Current image-based methods are based on either 2D-to-3D or 3D-to-2D find the correspondences, which ignore real scene geometric attributes. The main contribution of our paper that we use a 3D model reconstructed by short video as query realize 3D-to-3D under multi-task point retrieval framework. Firstly, enables us efficiently select location candidates. Furthermore, reconstruction exploits correlations among different...

10.1109/iccv.2015.280 article EN 2015-12-01

In this paper, we focus on the facial expression translation task and propose a novel Expression Conditional GAN (ECGAN) which can learn mapping from one image domain to another based an additional attribute. The proposed ECGAN is generic framework applicable different generation tasks where specific be easily controlled by conditional attribute label. Besides, introduce face mask loss reduce influence of background changing. Moreover, entire for recognition in wild, consists two modules,...

10.1109/icip.2019.8803654 article EN 2022 IEEE International Conference on Image Processing (ICIP) 2019-08-26

This paper present a low-cost wearable ultrasound bladder volume measurement and alarm system based on ARM9 DSP. The used new computing method to estimate the echo obtained by only one phased-array ultrasonic transducer. main signal processing task of are performed using DSP chip. estimated value was transmitted an Zigbee radio link. can be for continuous monitoring nursing unconsciousness elders, handicapped with spinal cord injury, other urological patients, children nocturnal enuresis...

10.1109/icbbe.2011.5781498 article EN 2011-05-01

Indoor localization has attracted a large amount of applications in mobile and robotics area, especially vast sophisticated environments. Most indoor methods are based on cellular base stations WiFi signals. Such require users to carry additional equipment. Localization accuracy is largely the beacon distribution. Image-based mainly applied for outdoor environments overcome problem caused by weak GPS signals building areas. In this paper, we propose localize images from multi-view settings....

10.5244/c.28.125 article EN 2014-01-01

The selection of discriminative features is an important and effective technique for many multimedia tasks. Using irrelevant in classification or clustering tasks could deteriorate the performance. Thus, designing efficient feature algorithms to remove a possible way improve With successful usage sparse models image video understanding, imposing structural sparsity \emph{feature selection} has been widely investigated during past years. Motivated by merit models, we propose novel method...

10.1145/2502081.2502142 article EN 2013-10-21

We present PET- the Pascal animal classes Eye Tracking database. Our database comprises eye movement recordings compiled from forty users for bird, cat, cow, dog, horse and sheep trainval sets VOC 2012 image set. Different recent eye-tracking databases such as [1, 2], a salient aspect of PET is that it contains movements recorded both free-viewing visual search task conditions. While some differences in terms overall gaze behavior scanning patterns are observed between two conditions, very...

10.1109/icme.2015.7177450 article EN 2022 IEEE International Conference on Multimedia and Expo (ICME) 2015-06-01
Coming Soon ...