LU Huimin

写真a

Title

Associate Professor

Laboratory

1-1 Sensui-cho, Tobata-ku, Kitakyushu-shi, Fukuoka

Research Fields, Keywords

Artificial Intelligence, Robotics, Oceanic Optics, Computer Vision

Homepage

http://www.ericlab.org

Degree 【 display / non-display

  • Kyushu Institute of Technology -  Doctor of Engineering  2014.03

Biography in Kyutech 【 display / non-display

  • 2019.09
    -
    Now

    Kyushu Institute of TechnologyFaculty of Engineering   Department of Mechanical and Control Engineering   Associate Professor  

Academic Society Memberships 【 display / non-display

  • 2019.12
    -
    Now
     

    IPSJ  JAPAN

Specialized Field (scientific research fund) 【 display / non-display

  • Perceptual information processing

  • Measurement engineering

 

Publications (Article) 【 display / non-display

  • Deep hierarchical encoding model for sentence semantic matching

    Lu W., Zhang X., Lu H., Li F.

    Journal of Visual Communication and Image Representation    71   2020.08  [Refereed]

     View Summary

    © 2020 Elsevier Inc. Sentence semantic matching (SSM) always plays a critical role in natural language processing. Measuring the intrinsic semantic similarity among sentences is very challenging and has not been substantially addressed. The latest SSM research usually relies on a shallow text representation and interaction between sentence pairs, which might not be enough to capture the complex semantic features and lead to limited performance. To capture more semantic context features and interactions, we propose a hierarchical encoding model (HEM) for sentence representation, further enhanced by a hierarchical matching mechanism for sentence interaction. Given two sentences, HEM generates intermediate and final representations in encoding layer, which are further handled by a novel hierarchical matching mechanism to capture more multi-view interactions in matching layer. The comprehensive experiments demonstrate that our model is capable to capture more sentence semantic features and interactions, which significantly outperforms the existing state-of-the-art neural models on the public real-world dataset.

    DOI Scopus

  • Dynamics and Isotropic Control of Parallel Mechanisms for Vibration Isolation

    Yang X., Wu H., Li Y., Kang S., Chen B., Lu H., Lee C.K.M., Ji P.

    IEEE/ASME Transactions on Mechatronics    25 ( 4 ) 2027 - 2034   2020.08  [Refereed]

     View Summary

    © 1996-2012 IEEE. Parallel mechanisms have been employed as architectures of high-precision vibration isolation systems. However, their performances in all degrees of freedom (DOFs) are nonidentical. The conventional solution to this problem is isotropic mechanism design, which is laborious and can hardly be achieved. This article proposes a novel concept; namely, isotropic control, to solve this problem. Dynamic equations of parallel mechanisms with base excitation are established and analyzed. An isotropic control framework is then synthesized in modal space. We derive an explicit relationship between modal control force and actuation force in joint space, enabling implementation of the isotropic controller. The multi-DOF system is transformed into multiidentical single-DOF systems. Under the framework of isotropic control, parallel mechanisms obtain an identical frequency response for all modes. An identical corner frequency, active damping, and rate of low-frequency transmissibility are achieved for all modes using a combining proportional, integral, and double integral compensator as a subcontroller. A 6-UPS parallel mechanism is presented as an example to demonstrate effectiveness of the new approach.

    DOI Scopus

  • Correlated Features Synthesis and Alignment for Zero-shot Cross-modal Retrieval

    Xu X., Lin K., Lu H., Gao L., Shen H.T.

    SIGIR 2020 - Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval      1419 - 1428   2020.07  [Refereed]

     View Summary

    © 2020 ACM. The goal of cross-modal retrieval is to search for semantically similar instances in one modality by using a query from another modality. Existing approaches mainly consider the standard scenario that requires the source set for training and the target set for testing share the same scope of classes. However, they may not generalize well on zero-shot cross-modal retrieval (ZS-CMR) task, where the target set contains unseen classes that are disjoint with the seen classes in the source set. This task is more challenging due to 1) the absence of the unseen classes during training, 2) inconsistent semantics across seen and unseen classes, and 3) the heterogeneous multimodal distributions between the source and target set. To address these issues, we propose a novel Correlated Feature Synthesis and Alignment (CFSA) approach to integrate multimodal feature synthesis, common space learning and knowledge transfer for ZS-CMR. Our CFSA first utilizes class-level word embeddings to guide two coupled Wassertein generative adversarial networks (WGANs) to synthesize sufficient multimodal features with semantic correlation for stable training. Then the synthetic and true multimodal features are jointly mapped to a common semantic space via an effective distribution alignment scheme, where the cross-modal correlations of different semantic features are captured and the knowledge can be transferred to the unseen classes under the cycle-consistency constraint. Experiments on four benchmark datasets for image-text retrieval and two large-scale datasets for image-sketch retrieval show the remarkable improvements achieved by our CFAS method comparing with a bundle of state-of-the-art approaches.

    DOI Scopus

  • Deep-Sea Organisms Tracking Using Dehazing and Deep Learning

    Lu H., Uemura T., Wang D., Zhu J., Huang Z., Kim H.

    Mobile Networks and Applications    25 ( 3 ) 1008 - 1015   2020.06  [Refereed]

     View Summary

    © 2018, Springer Science+Business Media, LLC, part of Springer Nature. Deep-sea organism automatic tracking has rarely been studied because of a lack of training data. However, it is extremely important for underwater robots to recognize and to predict the behavior of organisms. In this paper, we first develop a method for underwater real-time recognition and tracking of multi-objects, which we call “You Only Look Once: YOLO”. This method provides us with a very fast and accurate tracker. At first, we remove the haze, which is caused by the turbidity of the water from a captured image. After that, we apply YOLO to allow recognition and tracking of marine organisms, which include shrimp, squid, crab and shark. The experiments demonstrate that our developed system shows satisfactory performance.

    DOI Scopus

  • Learning adaptive contrast combinations for visual saliency detection

    Zhou Q., Cheng J., Lu H., Fan Y., Zhang S., Wu X., Zheng B., Ou W., Latecki L.J.

    Multimedia Tools and Applications    79 ( 21-22 ) 14419 - 14447   2020.06  [Refereed]

     View Summary

    © 2018, Springer Science+Business Media, LLC, part of Springer Nature. Visual saliency detection plays a significant role in the fields of computer vision. In this paper, we introduce a novel saliency detection method based on weighted linear multiple kernel learning (WLMKL) framework, which is able to adaptively combine different contrast measurements in a supervised manner. As most influential factor is contrast operation in bottom-up visual saliency, an average weighted corner-surround contrast (AWCSC) is first designed to measure local visual saliency. Combined with common-used center-surrounding contrast (CESC) and global contrast (GC), three types of contrast operations are fed into our WLMKL framework to produce the final saliency map. We show that the assigned weights for each contrast feature maps are always normalized in our WLMKL formulation. In addition, the proposed approach benefits from the advantages of the contribution of each individual contrast feature maps, yielding more robust and accurate saliency maps. We evaluated our method for two main visual saliency detection tasks: human fixed eye prediction and salient object detection. The extensive experimental results show the effectiveness of the proposed model, and demonstrate the integration is superior than individual subcomponent.

    DOI Scopus

display all >>

Publications (Books) 【 display / non-display

  • Cognitive Internet of Things: Frameworks, Tools and Applications

    ( Single Work )

    2020.01 ISBN: 978-3-030-04945-4

  • Artificial Inteligence and Robotics

    Huimin Lu, Xing Xu ( Joint Editor )

    Springer  2017.12 ISBN: 978-3-319-69877-9

  • Artificial Intelligence and Computer Vision

    Huimin Lu, Yujie Li ( Joint Editor )

    Springer  2016.11

Lectures 【 display / non-display

  • AIを活用した水中画像処理技術と深海資源調査への展開

    第1回海中海底工学フォーラム・ZERO   2019.04.12 

  • Artificial Intelligence in Deep-sea Observing

    The 2nd International Symposium on Artificial Intelligence and Robotics 2017   2017.11.25  ISAIR

  • Extreme Optical Imaging for Deep-sea Observing Network

    26th International Electrotechnical and Computer Science Conference ERK 2017   2017.09.25  IEEE Slovenia

  • Next Generation Artificial Intelligence in Society 5.0

    IBM Australia Seminar   2017.09.04  IBM Australia

 

Activities of Academic societies and Committees 【 display / non-display

  • 2019.08
    -
    Now

    IEEE Computer Society Big Data Special Technical Committee   Co-Chair