Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence最新文献

Extracting RoIs for Robust Far-infrared Pedestrian Detection on Board 鲁棒车载远红外行人检测的roi提取

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529494

Jianwei Zhang, Xiaoguang Yuan, Zhewei Xu, Wenjun Ke

{"title":"Extracting RoIs for Robust Far-infrared Pedestrian Detection on Board","authors":"Jianwei Zhang, Xiaoguang Yuan, Zhewei Xu, Wenjun Ke","doi":"10.1145/3529466.3529494","DOIUrl":"https://doi.org/10.1145/3529466.3529494","url":null,"abstract":"Far-infrared pedestrian detection onboard is more challenging compared to pedestrian detection of visible light. Existing works proved that the output of RoIs extraction, named as the proposal, is significantly related to recall rate and computational cost for pedestrian detection. However, it is non-trivial for RoIs extraction due to low resolution, blurred details, and pedestrian morphological features in far-infrared scenes. The paper proposes a novel RoIs extraction framework for far-infrared pedestrian detection on Board, named FIR-RoIEF, by using edge to obtain pedestrian contour feature and cascade filtering to gain valuable RoIs. Given pedestrian morphological features, we further present a vertical edge strategy to enhance pedestrian vertical features. A T-shaped template and RoIs reordering are used in the bounding box evaluation process to output pure and high-quality RoIs. Under basic and standard metrics, we perform experiments on public datasets, SCUT and KAIST, which both contain large far-infrared pedestrian objects. As for the result, given 100, 400, 1000 proposals, we both achieve the best recall in 1000 proposals, 94%, and 87% respectively. By the way, it can be proved that our method keeps a good balance between computational cost and good real-time.","PeriodicalId":375562,"journal":{"name":"Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131330059","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Controllable Person Image Synthesis GAN and Its Reconfigurable Energy-efficient Hardware Implementation 可控人体图像合成GAN及其可重构节能硬件实现

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529500

Shaoyue Lin, Yanjun Zhang

引用次数: 0

Human Action Recognition Based on Multi-Scale Feature Augmented Graph Convolutional Network 基于多尺度特征增强图卷积网络的人体动作识别

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529501

Wangyang Lv, Yinghua Zhou

引用次数: 1

Transformer-based Encoder-Decoder Model for Surface Defect Detection 基于变压器的表面缺陷检测编码器模型

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529471

Xiaofeng Lu, Wentao Fan

{"title":"Transformer-based Encoder-Decoder Model for Surface Defect Detection","authors":"Xiaofeng Lu, Wentao Fan","doi":"10.1145/3529466.3529471","DOIUrl":"https://doi.org/10.1145/3529466.3529471","url":null,"abstract":"Recently, deep learning approaches have been gaining popularity in industrial quality control (e.g. surface defect detection), due to their ability for automatically extracting more representative features. In this paper, we propose a two-stage end-to-end approach through a Transformer-based encoder-decoder for surface defect detection. First, we develop a surface defect detection model to train the slicing of input raw images with the same final resolution of the input images and the output images, which better expands the perceptual field. After that, a 1×1 convolution layer is applied to its final layer, thus reducing the number of channels to obtain a single-channel output mask. Then, we combine this single-channel output mask with the output obtained from the last layer of the first stage as the input of the second stage decision layer. Considering different types of sample data, we design two different decision network strategies, namely: plain-up sampling and dynamic-up sampling. Our experimental studies on several publicly available datasets show that the proposed approach is general and effective in detecting defects, and we only need a relatively small number of samples to train the model, which has a good applicability in industrial practice where the sample size is normally limited.","PeriodicalId":375562,"journal":{"name":"Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130253963","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 1

Mass Mahjong Decision System Based on Transfer Learning 基于迁移学习的群体麻将决策系统

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529485

Yajun Zheng, Shuqin Li

引用次数: 0

Neural Sign Language Translation with SF-Transformer 基于SF-Transformer的神经手语翻译

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529503

Qifang Yin, Wenqi Tao, Xiaolong Liu, Yu Hong

引用次数: 3

Object Detection Based on Feature Balance Pyramid in UAV Imagery 基于特征平衡金字塔的无人机图像目标检测

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529469

Jiao Xu, Jian Xu, Zeming Xu, Zhengguang Xie

{"title":"Object Detection Based on Feature Balance Pyramid in UAV Imagery","authors":"Jiao Xu, Jian Xu, Zeming Xu, Zhengguang Xie","doi":"10.1145/3529466.3529469","DOIUrl":"https://doi.org/10.1145/3529466.3529469","url":null,"abstract":"Compared with images taken from the ground perspective, small objects account for a large proportion of aerial UAV images and the image perspective changes greatly, which affects the target detection effect of aerial UAV images. In this paper, the Yolov5 algorithm is improved to adapt to UAV object detection. Given a large number of small objects in aerial images, the feature balance pyramid structure is added to improve the loss of low-level features and improve the detection effect of the small object. In the feature balance pyramid, Pixel un-Shuffle is used to adjust the scale of the feature, which preserves the low-level feature information and reduces the computational cost. The cross self-attention module is proposed to improve the balanced feature map and improve the positioning accuracy of the small object. The Angle of view of aerial images varies greatly. In this paper, the deformable convolutional network is added to the backbone network of Yolov5 to enhance the feature extraction capability of the model for multi-view objects. Experimental results show that on the visdrone data set, the improved algorithm improves the average accuracy (mAP) by 1.4 percentage points compared with the original algorithm.","PeriodicalId":375562,"journal":{"name":"Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131456561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Chinese Named Entity Recognition Fusing Lexical and Syntactic Information 融合词法和句法信息的中文命名实体识别

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529506

Min Zhang, Bicheng Li, Qilong Liu, Jing Wu

{"title":"Chinese Named Entity Recognition Fusing Lexical and Syntactic Information","authors":"Min Zhang, Bicheng Li, Qilong Liu, Jing Wu","doi":"10.1145/3529466.3529506","DOIUrl":"https://doi.org/10.1145/3529466.3529506","url":null,"abstract":"Chinese named entity recognition is an important research field in natural language processing and is a significant factor in multitasking. However, the accuracy of Chinese named entity recognition is impaired by the lack of entity boundary information and the neglect of potential syntactic information. In this context. In the present study, a Chinese named entity recognition fusing lexical and syntactic information was proposed, and the steps are as follows. Firstly, input text is mapped to character vectors, and external lexical information is introduced with an improved word set matching method and integrated into the input representation of each character. Secondly, BiLSTM is used to obtain the context vector based on the input representation of the character. Thirdly, the syntactic vector is constructed by means of the key value memory network, and the feature vector is obtained by the weighted fusion of context vector and syntactic vector by the gating mechanism. Finally, the feature vector is input into CRF to realize Chinese named entity recognition. The experimental results reveal that the proposed method had better performance on Resume, Weibo and MSRA, and was superior to the current mainstream Chinese named entity recognition methods.","PeriodicalId":375562,"journal":{"name":"Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132827773","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 1

Named Entity Recognition of Zhuang Language Based on the Feature of Initial Letter in Word 基于词首字母特征的壮语命名实体识别

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529478

Weiquan Zhang, Suqin Tang, Danni He, Tinghui Li, Changchun Pan

{"title":"Named Entity Recognition of Zhuang Language Based on the Feature of Initial Letter in Word","authors":"Weiquan Zhang, Suqin Tang, Danni He, Tinghui Li, Changchun Pan","doi":"10.1145/3529466.3529478","DOIUrl":"https://doi.org/10.1145/3529466.3529478","url":null,"abstract":"Named entity recognition is an important task and basis for the intelligent information processing and knowledge representation learning of the Zhuang Language. A BilSTM-CNN-CRF network model combining the uppercase and lowercase characters of words is proposed to be applied to the named entity recognition task of the Zhuang language, which lacks corpus for named entity labeling. Firstly, word2vec is used to train in unmarked Zhuang text to get the word vector of the Zhuang language. Then convolutional neural network is used to extract the character features of Zhuang words, and the character feature vector is obtained. The above two vectors were connected with the initial case feature vectors, which are randomly generated, and then the connected vectors were input into a BilSTM-CNN-CRF model for training; thus, the end-to-end named entity recognition model of Zhuang language was constructed. Experimental results show that, without relying on artificial features and external dictionaries, the proposed method in this study is superior to contrastive models by achieving an 80.37% F1 value in the named entity recognition task, which leads to the realization of automated named entity recognition of Zhuang language.","PeriodicalId":375562,"journal":{"name":"Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127709961","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Mixture Density Hyperspherical Generative Adversarial Networks 混合密度超球面生成对抗网络

Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence Pub Date : 2022-03-04 DOI: 10.1145/3529466.3529475

Qinyang Li, Wentao Fan

{"title":"Mixture Density Hyperspherical Generative Adversarial Networks","authors":"Qinyang Li, Wentao Fan","doi":"10.1145/3529466.3529475","DOIUrl":"https://doi.org/10.1145/3529466.3529475","url":null,"abstract":"The Generative Adversarial Networks (GANs) are deep generative models that can generate realistic samples, but they are difficult to train in practice due to the problem of mode collapse, where the generator only repeatedly generates one mode in samples during the learning process, or only generates a small number of modes after reaching the Nash equilibrium during the adversarial training. In order to solve this issue while making the generator contains promising generation ability, we propose a mixture density hyperspherical generative model namely MDH-GAN that combines variational autoencoder (VAE) and generative adversarial network. Unlike most of the GAN-based generative models that consider a Gaussian prior, MDH-GAN adopts the von Mises-Fisher (vMF) prior defined on a unit hypersphere. Our model combines VAE with GAN by integrating the encoder of VAE with GAN to form a jointly training framework. Therefore, the generator of our model can learn data distribution with a hyperspherical latent structure, leading to an improved generative ability of the generator. Moreover, a vMF mixture model is deployed in the discriminator to form a hypersphere space to avoid mode collapse of the model. In our experiments, by calculating the Fréchet Inception distance (FID) between the generated images and real ones, we prove that MDH-GAN has a better ability to generate high-quality images with high diversity.","PeriodicalId":375562,"journal":{"name":"Proceedings of the 2022 6th International Conference on Innovation in Artificial Intelligence","volume":"146 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123321311","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0