摘要:
In recent years, cross-modal hashing has attracted an increasing attention due to its fast retrieval speed and low storage requirements. However, labeled datasets are limited in real application, and existing unsupervised cross-modal hashing algorithms usually employ heuristic geometric prior as semantics, which introduces serious deviations as the similarity score from original features cannot reasonably represent the relationships among instances. In this paper, we study the unsupervised deep cross-modal hash retrieval method and propose a novel Semantic Graph Evolutionary Hashing (SGEH) to solve the above problem. The key novelty of SGEH is its evolutionary affinity graph construction method. To be concrete, we explore the sparse similarity graph with clustering results, which evolve from fusing the affinity information from code-driven graph on intrinsic data and subsequently extends to dense hybrid semantic graph which restricts the process of hash code learning to learn more discriminative results. Moreover, the batch-inputs are chosen from edge set rather than vertexes for better exploring the original spatial information in the sparse graph. Experiments on four benchmark datasets demonstrate the superiority of our framework over the state-of-the-art unsupervised cross-modal retrieval methods. Code is available at: https://github.com/theusernamealreadyexists/SGEH.
作者机构:
[Zhang, Cheng; Liu, Hai] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.;[Deng, Yongjian] Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China.;[Deng, Yongjian] Minist Educ, Engn Res Ctr Intelligence Percept & Autonomous Co, Beijing, Peoples R China.;[Li, Youfu; Xie, Bochen] City Univ Hong Kong, Dept Mech Engn, Kowloon, Hong Kong, Peoples R China.
会议名称:
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
会议时间:
JUN 17-24, 2023
会议地点:
Vancouver, CANADA
会议主办单位:
[Zhang, Cheng;Liu, Hai] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.^[Deng, Yongjian] Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China.^[Deng, Yongjian] Minist Educ, Engn Res Ctr Intelligence Percept & Autonomous Co, Beijing, Peoples R China.^[Xie, Bochen;Li, Youfu] City Univ Hong Kong, Dept Mech Engn, Kowloon, Hong Kong, Peoples R China.
会议论文集名称:
IEEE Conference on Computer Vision and Pattern Recognition
摘要:
Head pose estimation (HPE) has been widely used in the fields of human machine interaction, self-driving, and attention estimation. However, existing methods cannot deal with extreme head pose randomness and serious occlusions. To address these challenges, we identify three cues from head images, namely, neighborhood similarities, significant facial changes, and critical minority relationships. To leverage the observed findings, we propose a novel critical minority relationship-aware method based on the Transformer architecture in which the facial part relationships can be learned. Specifically, we design several orientation tokens to explicitly encode the basic orientation regions. Meanwhile, a novel token guide multiloss function is designed to guide the orientation tokens as they learn the desired regional similarities and relationships. We evaluate the proposed method on three challenging benchmark HPE datasets. Experiments show that our method achieves better performance compared with state-of-the-art methods. Our code is publicly available at https://github.com/zc2023/TokenHPE.
作者机构:
[Deng, Yongjian] Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China.;[Chen, Hao] Southeast Univ, Sch Comp Sci & Engn, Nanjing, Peoples R China.;[Liu, Hai] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Hubei, Peoples R China.;[Li, Youfu] City Univ Hong Kong, Dept Mech Engn, Hong Kong, Peoples R China.
会议名称:
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
会议时间:
JUN 18-24, 2022
会议地点:
New Orleans, LA
会议主办单位:
[Deng, Yongjian] Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China.^[Chen, Hao] Southeast Univ, Sch Comp Sci & Engn, Nanjing, Peoples R China.^[Liu, Hai] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Hubei, Peoples R China.^[Li, Youfu] City Univ Hong Kong, Dept Mech Engn, Hong Kong, Peoples R China.
会议论文集名称:
IEEE Conference on Computer Vision and Pattern Recognition
摘要:
Event cameras attract researchers' attention due to their low power consumption, high dynamic range, and extremely high temporal resolution. Learning models on event-based object classification have recently achieved massive success by accumulating sparse events into dense frames to apply traditional 2D learning methods. Yet, these approaches necessitate heavy-weight models and are with high computational complexity due to the redundant information introduced by the sparse-to-dense conversion, limiting the potential of event cameras on real-life applications. This study aims to address the core problem of balancing accuracy and model complexity for event-based classification models. To this end, we introduce a novel graph representation for event data to exploit their sparsity better and customize a lightweight voxel graph convolutional neural network (EV-VGCNN) for event-based classification. Specifically, (1) using voxel-wise vertices rather than previous point-wise inputs to explicitly exploit regional 2D semantics of event streams while keeping the sparsity; (2) proposing a multi-scale feature relational layer (MFRL) to extract spatial and motion cues from each vertex discriminatively concerning its distances to neighbors. Comprehensive experiments show that our model can advance state-of-the-art classification accuracy with extremely low model complexity (merely 0.84M parameters).
作者机构:
[Zhang, H; Tong, Hang; Liu, Sanya; Li, Yaopeng; Zhang, Hao; Min, Yuandong] Cent China Normal Univ, Fac Artificial Intelligence Educ, Natl Engn Res Ctr Learning, Wuhan 430079, Peoples R China.;[Zhang, H; Tong, Hang; Liu, Sanya; Li, Yaopeng; Zhang, Hao; Min, Yuandong] Cent China Normal Univ, Fac Artificial Intelligence Educ, Natl Engn Lab Educ Big Data, Wuhan 430079, Peoples R China.
会议名称:
IEEE International Performance, Computing, and Communications Conference (IPCCC)
会议时间:
NOV 11-13, 2022
会议地点:
Austin, TX
会议主办单位:
[Zhang, Hao;Min, Yuandong;Liu, Sanya;Tong, Hang;Li, Yaopeng] Cent China Normal Univ, Fac Artificial Intelligence Educ, Natl Engn Res Ctr Learning, Wuhan 430079, Peoples R China.^[Zhang, Hao;Min, Yuandong;Liu, Sanya;Tong, Hang;Li, Yaopeng] Cent China Normal Univ, Fac Artificial Intelligence Educ, Natl Engn Lab Educ Big Data, Wuhan 430079, Peoples R China.
会议论文集名称:
IEEE International Performance Computing and Communications Conference (IPCCC)
摘要:
Event cameras asynchronously capture pixel-level intensity changes in scenes and output a stream of events. Compared with traditional frame-based cameras, they can offer competitive imaging characteristics: low latency, high dynamic range, and low power consumption. It means that event cameras are ideal for vision tasks in dynamic scenarios, such as human action recognition. The best-performing event-based algorithms convert events into frame-based representations and feed them into existing learning models. However, generating informative frames for long-duration event streams is still a challenge since event cameras work asynchronously without a fixed frame rate. In this work, we propose a novel frame-based representation named Compact Event Image (CEI) for action recognition. This representation is generated by a self-attention based module named Event Tubelet Compressor (EVTC) in a learnable way. The EVTC module adaptively summarizes the long-term dynamics and temporal patterns of events into a CEI frame set. We can combine EVTC with conventional video backbones for end-to-end event-based action recognition. We evaluate our approach on three benchmark datasets, and experimental results show it outperforms state-of-the-art methods by a large margin.
作者机构:
[Zhang, Kui; Dai, Zhicheng; Wang, Chunran; Chen, Rongjin; Zhu, Fuming] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
会议名称:
10th International Conference on Information and Education Technology (ICIET)
会议时间:
APR 09-11, 2022
会议地点:
Matsue, JAPAN
会议主办单位:
[Dai, Zhicheng;Zhang, Kui;Wang, Chunran;Chen, Rongjin;Zhu, Fuming] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
摘要:
It is an important issue to effectively perceive learners' emotional state in the field of smart education, which helps to enhance the interaction between teaching groups and stimulate learners' enthusiasm for learning. Taking "emotion perception" as the theme, this paper retrieved 533 relevant core literature from the CNKI (China National Knowledge Infrastructure) database and used econometric methods and visualization software CiteSpace to analyze the number of literature, authors, institutions, and keywords. The results show that the number of literature published on learners' emotion perception has increased year by year in the past 30 years and is in the mature stage of development; The authors and institutions are relatively scattered, and the core research system of emotion perception has not been formed. Building a multimodal perception model using techniques such as expression recognition, posture recognition, and physiological parameter detection is research hotspots in the field of emotion perception. The research trends in this field are to collect and fuse multimodal emotional data, and deeply analyze the change rule of learners' emotions based on deep learning and data mining.
作者机构:
[Chen, Meng; Xu, Jian; Wu, Chen; Ma, Jiman; Wu, Di] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.;[Ma, Binbin] South Cent Univ Nationalities, Sch Management, Wuhan, Peoples R China.
会议名称:
7th International Conference of the Immersive-Learning-Research-Network (iLRN)
会议时间:
MAY 17-JUN 10, 2021
会议地点:
ELECTR NETWORK
会议主办单位:
[Wu, Chen;Chen, Meng;Wu, Di;Ma, Jiman;Xu, Jian] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.^[Ma, Binbin] South Cent Univ Nationalities, Sch Management, Wuhan, Peoples R China.
关键词:
ICT evaluation process in education;real-time monitoring;3D visualization;CesiumJS
摘要:
ICT (information and communications technology) evaluation in education is a key step in measuring the development level of education informatization. At present, the statistical data of evaluation process mainly be transmitted in the form of static chart files. How to grasp the overall evaluation status of each evaluation area in real time and understand the latest progress and bottlenecks of evaluation work is an urgent problem. This work-in-progress paper proposes a real-time monitoring system for ICT evaluation process in education based on CesiumJS 3D visualization, which can provide multi-modal 3D visualization for dynamic evaluation process data and further support knowledge mining.
作者机构:
[Li, Qing; Sun, Jianwen] Cent China Normal Univ, Natl Engn Lab Educ Big Data, Wuhan, Peoples R China.;[Lu, Zijian; Zhou, Jianpeng; Zhang, Kai] Cent China Normal Univ, Natl Engn Res Ctr Learning, Wuhan, Peoples R China.
会议名称:
14th International Conference on Knowledge Science, Engineering, and Management (KSEM)
会议时间:
AUG 14-16, 2021
会议地点:
Tokyo, JAPAN
会议主办单位:
[Sun, Jianwen;Li, Qing] Cent China Normal Univ, Natl Engn Lab Educ Big Data, Wuhan, Peoples R China.^[Zhou, Jianpeng;Zhang, Kai;Lu, Zijian] Cent China Normal Univ, Natl Engn Res Ctr Learning, Wuhan, Peoples R China.
摘要:
Knowledge tracing predicts students' future performance based on their past performance. Most of the existing models take skills as input, which neglects question information and further limits the model performance. Inspired by item-item collaborative filtering in recommender systems, we propose a question-question Collaborative embedding method for Knowledge Tracing (CoKT) to introduce question information. To be specific, we incorporate student-question interactions and question-skill relations to capture question similarity. Based on the similarity, we further learn question embeddings, which are then integrated into a neural network to make predictions. Experiments demonstrate that CoKT significantly outperforms baselines on three benchmark datasets. Moreover, visualization illustrates that CoKT can learn interpretable question embeddings and achieve more obvious improvement on AUC when the interaction data is more sparse.
摘要:
Graph Convolutional Networks (GCN) and their variants have achieved brilliant results in graph representation learning. However, most existing methods cannot be utilized for deep architectures and can only capture the low order proximity in networks. In this paper, we have proposed a Residual Simple Graph Convolutional Network (RSGCN), which can aggregate information from distant neighbor node features without over-smoothing and vanishing gradients. Given that node features of the same class have certain similarity, a weighted feature propagation is considered to ensure effective information aggregation by giving higher weights to similar neighbor nodes. Experimental results on several datasets of node classification demonstrate the proposed methods outperform the state-of-the-art methods in terms of effectiveness and efficiency.
作者机构:
[Yu, Xin Guo; He, Bin; Zhuang, Jiao Jiao; Sun, Jia Yu; Dai, Zi Chun] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
通讯机构:
[Yu, Xin Guo] C;Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
会议名称:
3rd International Conference on Mechatronics and Intelligent Robotics (ICMIR)
会议时间:
MAY 25-26, 2019
会议地点:
Kunming, PEOPLES R CHINA
会议主办单位:
[Yu, Xin Guo;Sun, Jia Yu;He, Bin;Zhuang, Jiao Jiao;Dai, Zi Chun] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
摘要:
This paper designs and implements the automatic invigilation functions using the embedded technology. It proposes a framework for automatic invigilation, which conducts the invigilation functions of the entire examination process. In the examination preparation stage, the framework collects the registration details of examinees and verifies the details with the database through the remote server. During the examination ongoing stage, it keeps checking the consistence between each examinee and his examination materials by photographing the examinee and scanning the QR codes on examination papers, sketch papers, and answer sheets. In the examination ending stage, it checks the consistence of an examination bag and the materials being put into it by scanning and verifying the QR codes on them. The framework reduces the human workload by using automatic functions to replace the human work. The tests demonstrates that the framework can perform the designed functions. (C) 2020 The Authors. Published by Elsevier B.V.
作者机构:
[Huo, Jiao] Wuhan Zhimei Technol Co Ltd, Wuhan, Peoples R China.;[Liu, Leyuan; Zhao, Yi] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
会议名称:
11th International Symposium on Multispectral Image Processing and Pattern Recognition (MIPPR) - Pattern Recognition and Computer Vision
会议时间:
NOV 02-03, 2019
会议地点:
Wuhan, PEOPLES R CHINA
会议主办单位:
[Zhao, Yi;Liu, Leyuan] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.^[Huo, Jiao] Wuhan Zhimei Technol Co Ltd, Wuhan, Peoples R China.
摘要:
Nowadays, more and more printed books are accompanied by electronic resources including videos, audios, games, augmented reality and other mobile apps. However, it is not very convenient to access most of these electronic resources, as the association between printed books and electronic resources is not automatically available [2]. To build a bridge between a book page and the corresponding electronic resources, a large-scale book page retrieval method using deep hashing network is presented in this paper. There are mainly three contributions: First, a pipeline is proposed to make a Convolutional Neural Network (CNN) trained for another unrelated task available for book page retrieval. Second, the high-dimensional features extracted from the CNN is mapped to the low-dimensional binary hash code sequence in Hemming space by the deep hashing network, which not only increases the speed of retrieval but also saves the space of feature storage. Third, a large-scale dataset which is consist of 1.55M book page images is collected. Experimental results on the 1.55M book page dataset show that the proposed deep hashing network achieves a Top-1 hit rate of 92.1% and the response time is less than 0.6 second on a desktop computer with a GeForce 1080Ti GPU.
摘要:
Synthesizing high-resolution realistic images from text description using one iteration Generative Adversarial Network (GAN) is difficult without using any additional techniques because mostly the blurry artifacts and mode collapse problems are occurring. To reduce these problems, this paper proposes an Iterative Generative Adversarial Network (iGAN) which takes three iterations to synthesize high-resolution realistic image from their text description. In the \(1^{st}\) iteration, GAN synthesizes a low-resolution \(64 \times 64\) pixels basic shape and basic color image from the text description with less mode collapse and blurry artifacts problems. In the \(2^{nd}\) iteration, GAN takes the result of the \(1^{st}\) iteration and text description again and synthesizes a better resolution \(128 \times 128\) pixels better shape and well color image with very less mode collapse and blurry artifacts problems. In the last iteration, GAN takes the result of the \(2^{nd}\) iteration and text description as well and synthesizes a high-resolution \(256 \times 256\) well shape and clear image with almost no mode collapse and blurry artifacts problems. Our proposed iGAN shows a significant performance on CUB birds and Oxford-102 flowers datasets. Moreover, iGAN improves the inception score and human rank as compare to the other state-of-the-art methods.
摘要:
The linear canonical transforms (LCTs) are a Lie group of transforms including the Fresnel and Fourier transforms that describe scalar wave propagation in quadratic phase systems. As such, they are useful in system analysis and design, and their discretisations are important for opto-numerical systems, e.g. numerical reconstruction algorithms in digital holography. An important topic in the literature is therefore the generalization of Fourier transform properties for the LCTs. A number of authors have proposed convolution theorems for the linear canonical transform, with different goals in mind. In this paper, we compare those methods, with particular attention being paid to the consequences of discretization. In a similar way to how discrete convolution associated with the DFT differs from that associated with the Fourier transform, we must take the chirp-periodic nature of discrete LCTs into account when determining the discrete convolution associated with LCTs. This work is of significance for the simulation of VanderLugt correlators, which have been used for optical implementations of neural networks, and for optical filtering operations and coherent optical signal processing in general.
摘要:
With the further application of information technology in the field of education, the education informationization gradually moves from shallow application to deep integration. The promotion of education informationization makes the integration problem more complicated, so it is urgent to study and discuss how to promote the deepening of the application of information instruction in school. Based on the view of education ecology, this paper puts forward a teaching model to promote the deep integration of information technology in teaching, and to establish an information-based teaching ecological model, and also puts forward the idea to reconstruct the school teaching ecology from the concept layer, pattern layer and practice layer. At the same time, combined with the examples of middle school information teaching ecology remodeling, this paper expounds how to optimize and reform the teaching ways, organization management mode and campus culture through teaching model construction.
作者机构:
[Wu, Di; Zhou, Chi; Shi, Yinghui; Chen, Min; Yang, Wei] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan, Peoples R China.
会议名称:
8th International Conference on Educational Innovation through Technology (EITT)
会议时间:
OCT 27-31, 2019
会议地点:
Univ Southern Mississippi, Biloxi, MS
会议主办单位:
Univ Southern Mississippi
会议论文集名称:
Proceedings of the International Conference of Educational Innovation through Technology
关键词:
technology integration;mathematic teachers;Information and Communication Technology (ICT) application;education informationization;differences
摘要:
The application of Information and Communication Technology (ICT) in teaching can improve the teaching quality and efficiency. Exploring the differences in mathematic teachers' ICT application level can not only better help us understand the status of overall mathematic teachers' ICT application level, but also promote the development of ICT-based teaching in mathematic by discovering the shortcomings. This study explored the differences of K-12 mathematic teachers' ICT application level from three aspects-teacher's attitude towards ICT (TAT), ICT instruction in classroom (IIC) and ICT effects (ICE). A survey research design was used for the study, and 918 K-12 mathematic teachers participated in the study. The ANOVA results showed that, there were significant differences in TAT and ICE between primary school teachers and secondary school teachers, while there were no significant differences between rural school teachers and urban school teachers. As for IIC, on the other hand, there was significant differences between primary school teachers and secondary school teachers and there were significant differences between rural school teachers and urban school teachers. According to the results, some implications were proposed to improve K-12 mathematic teachers' ICT application level.
作者机构:
[Mei, Yunshan; Liu, Sanya; Zhang, Hao; Huang, Tao; Yang, Huali] Cent China Normal Univ, Natl Engn Lab Educ Big Data, Wuhan 430079, Hubei, Peoples R China.;[Mei, Yunshan; Liu, Sanya; Zhang, Hao; Huang, Tao; Yang, Huali] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan 430079, Hubei, Peoples R China.
会议名称:
IEEE 9th International Conference on Electronics Information and Emergency Communication (ICEIEC)
会议时间:
JUL 12-14, 2019
会议地点:
Beijing, PEOPLES R CHINA
会议主办单位:
[Huang, Tao;Mei, Yunshan;Zhang, Hao;Liu, Sanya;Yang, Huali] Cent China Normal Univ, Natl Engn Lab Educ Big Data, Wuhan 430079, Hubei, Peoples R China.^[Huang, Tao;Mei, Yunshan;Zhang, Hao;Liu, Sanya;Yang, Huali] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Wuhan 430079, Hubei, Peoples R China.
会议论文集名称:
IEEE International Conference on Electronics Information and Emergency Communication
摘要:
Engagement is an important measure of users' learning experience in online learning environment. Improving the accuracy of engagement recognition can help the instructors get timely feedback on the courses, optimize the recommendation strategies of online learning platforms and enhance users' learning experience. In this paper, we propose a novel model: Deep Engagement Recognition Network (DERN) which combines temporal convolution, bidirectional LSTM and attention mechanism to identify the degree of engagement based on the features captured by OpenFace. In order to verify the validity and stability of the model, we evaluate the accuracy by the way of five-fold cross-validation. Finally, we achieved 60% in top-1 accuracy in the problem of four classification for engagement on the dataset called DAISEE which provided a baseline of 57.9%.
作者机构:
[Wang, Yufan] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Hubei Prov Key Lab Artificial Intelligence & Smar, Natl Language Resources Monitor & Res Ctr Network, Wuhan, Peoples R China.;[Fan, Rui; Tu, Xinhui; He, Tingting; Zhou, Wenji] Cent China Normal Univ, Sch Comp, Hubei Prov Key Lab Artificial Intelligence & Smar, Natl Language Resources Monitor & Res Ctr Network, Wuhan, Peoples R China.
会议名称:
IEEE International Conference on Big Data (Big Data)
会议时间:
DEC 09-12, 2019
会议地点:
Los Angeles, CA
会议主办单位:
[Wang, Yufan] Cent China Normal Univ, Natl Engn Res Ctr E Learning, Hubei Prov Key Lab Artificial Intelligence & Smar, Natl Language Resources Monitor & Res Ctr Network, Wuhan, Peoples R China.^[He, Tingting;Fan, Rui;Zhou, Wenji;Tu, Xinhui] Cent China Normal Univ, Sch Comp, Hubei Prov Key Lab Artificial Intelligence & Smar, Natl Language Resources Monitor & Res Ctr Network, Wuhan, Peoples R China.
会议论文集名称:
IEEE International Conference on Big Data
关键词:
external knowledge;intent detection;slot filling;spoken language understanding
摘要:
At present, spoken language understanding (SLU) in multi-turn dialogue is a research hotspot, which mainly includes intent detection and slot filling. SLU models trained by large-scale corpus can learn good superficial semantic and grammatical information. But they lack the ability for modeling the knowledge needed to understand language. In order to further understand the deep semantic information of the dialogue, external knowledge needs to be modeled and incorporated into the SLU model. In addition, utilizing the correlation between history dialogue and current utterance is able to understand dialogue in multi-turn SLU. Thus, this paper proposes a joint model of intent detection and slot filling based on history context and external knowledge. This model constructs history dialogue encoder to obtain history context. Meanwhile, it constructs knowledge attention over context module. This module selects external knowledge according to the context information in current utterance and obtains knowledge representation. Finally, the history context and external knowledge representation are incorporated into the intent detection and slot filling joint model. The result of experiments on the common dataset demonstrate that with the help of external knowledge and history context, the performance of our model has a significant improvement.