Proceedings of the 2022 5th International Conference on Image and Graphics Processing最新文献_第2页

Research on Handwritten Digital Image Recognition Model Based On Deep Learning and Construction of Browser Service Platform 基于深度学习的手写数字图像识别模型研究及浏览器服务平台的构建

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512404

Han-Ting Huang, Zhu Chen, Tongyuan Bai, Zhihong Zhao

引用次数: 0

Image Inpainting Based on Edge Features and Attention Mechanism 基于边缘特征和注意机制的图像补图

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512398

Yuting Fu, Dan Xu, Kangjian He, Haipeng Li, Tingting Zhang

引用次数: 0

Tiny Object Detection based on YOLOv5 基于YOLOv5的微小目标检测

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512395

Tongyuan Huang, Minhao Cheng, Yuling Yang, Xiangling Lv, Jia Xu

{"title":"Tiny Object Detection based on YOLOv5","authors":"Tongyuan Huang, Minhao Cheng, Yuling Yang, Xiangling Lv, Jia Xu","doi":"10.1145/3512388.3512395","DOIUrl":"https://doi.org/10.1145/3512388.3512395","url":null,"abstract":"In view of the poor accuracy of mainstream object detection algorithms in detecting tiny objects, A tiny object detection algorithm based on improved YOLOv5 is proposed. The main feature extraction network of YOLOv5 was modified to generate four feature images to enhance feature extraction of the original input images. Modified the YOLOv5 Neck part, combined with FPN and PANet, carried out feature fusion for four feature maps containing different semantic information, generated better features, and improved the performance of tiny object detection. GIoU loss function was introduced to replace the IoU loss function in the original algorithm to improve the positioning accuracy of tiny objects. Swish activation function was used to replace the original ReLU activation function to better retain target features. The Mosaic data enhancement method was used to enrich the object detection background, and the learning rate cosine annealing attenuation training method was used to dynamically update the learning rate parameters, and the improved YoloV5 algorithm was fused. In this paper, a comparison test is conducted between the original YoloV5 algorithm and CityPrersons data set. Experimental results show that the improved YoloV5 algorithm can effectively improve the detection accuracy of tiny objects.","PeriodicalId":434878,"journal":{"name":"Proceedings of the 2022 5th International Conference on Image and Graphics Processing","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-01-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132380598","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: 3

Application of Image Similarity Detection based on Typical ITAI Servers 基于典型ITAI服务器的图像相似度检测应用

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512392

Su-Jiau Chen, Zhimin Wu, M. Guo, Zhenyu Wang

引用次数: 0

Using DSCB: A Depthwise Separable Convolution Block Rebuild MTCNN for Face Detection 基于深度可分离卷积块重建MTCNN人脸检测

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512389

Qiang Wang, Jingru Cui, Zunying Qin, Ninggang An, Xiaofei Ma, Guodong Li

{"title":"Using DSCB: A Depthwise Separable Convolution Block Rebuild MTCNN for Face Detection","authors":"Qiang Wang, Jingru Cui, Zunying Qin, Ninggang An, Xiaofei Ma, Guodong Li","doi":"10.1145/3512388.3512389","DOIUrl":"https://doi.org/10.1145/3512388.3512389","url":null,"abstract":"Nowadays, there are huge demands of face detection in images and videos for surveillance, education, autonomous driving and health care. These application scenarios need high accuracy and efficiency of face detection. However, in some scene, unconstrained pose variation, occlusion, large number of faces and illumination bring great challenges to existing face detection methods. In view of above problems, we propose a depthwise separable convolution block (DSCB) which can maintain the speed of training and improve the accuracy at the same time. Then, using the proposed DSCB, we design a face detection model based on MTCNN (Multi-task Convolution Neural Network) to improve performance of occlusion, unconstrained pose variation, large numbers of small targets. In order to better evaluate the proposed method, we built a new dataset which is derived from the classroom teaching scene for training and evaluating. Our dataset consists of 7168 images and 294924 face bounding boxes with occlusion, unconstrained pose variation, and large numbers of small targets. The comparative experiments on our dataset show that the proposed method is superior to other state-of-the-art methods in accuracy and speed of face detection. Compared with the original MTCNN, the face detection method we proposed can bring about 3.9%, 8.66% and 1.39 times overall performance improvement on precision, recall and detection speed respectively.","PeriodicalId":434878,"journal":{"name":"Proceedings of the 2022 5th International Conference on Image and Graphics Processing","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-01-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131677874","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

Bed-Leaving Action Recognition Based on YOLOv3 and AlphaPose 基于YOLOv3和AlphaPose的离床动作识别

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512406

Caixia Zhang, Xiaoyu Yang

引用次数: 2

Anatomy-guided Multi-View Fusion Framework for Abdominal CT Multi-Organ Segmentation 解剖引导下腹部CT多器官分割的多视点融合框架

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512413

Zhongwei Yang, Haopeng Kuang, Xukun Zhang, Yang Liu, Peng Zhai, Lubin Chen, Lihua Zhang

引用次数: 1

Structure-based Street Tree Extraction from Mobile Laser Scanning Point Clouds 基于结构的移动激光扫描点云街道树提取

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512443

W. Hao, Zhanbin Zuo, W. Liang

引用次数: 0

A Vision-based Monitoring System for Quality Assessment of Fused Filament Fabrication (FFF) 3D Printing 基于视觉的熔丝制造(FFF) 3D打印质量评估监测系统

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512424

Jingdong Li, Wei Quan, L. Shark, H. Brooks

{"title":"A Vision-based Monitoring System for Quality Assessment of Fused Filament Fabrication (FFF) 3D Printing","authors":"Jingdong Li, Wei Quan, L. Shark, H. Brooks","doi":"10.1145/3512388.3512424","DOIUrl":"https://doi.org/10.1145/3512388.3512424","url":null,"abstract":"As one of the most popular 3D printing technology, Fused Filament Fabrication (FFF) allows intricate structures to be produced without complex manufacturing processes. However, there is a limitation of the currently available FFF 3D printers which print blindly without an ability to detect and stop upon printing deviations, incurring additional running costs due to unnecessary waste of materials and time. This has led to a novel development reported in this paper of a vision-based monitoring system for the quality assessment of 3D printing by applying advanced computer vision algorithms and imaging processing techniques. The proposed approach is through comparison between actual images of the printed layer and simulated images created by slicing CAD model via G-code generation based on the calibrated camera pose. Also presented are feature extraction methods to yield object dimension, profile and infill for quality assessment, with the system performance demonstrated based on various object geometries. Using this system makes it possible to analyze and examine the quality of 3D printing during the print process, which could identify the defective printed parts, terminate the whole process and alert the users for time and cost-savings.","PeriodicalId":434878,"journal":{"name":"Proceedings of the 2022 5th International Conference on Image and Graphics Processing","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-01-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116794128","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

Consistency Mean-Teaching for Unsupervised Domain Adaptive Person Re-identification 无监督领域自适应人再识别的一致性方法教学

Proceedings of the 2022 5th International Conference on Image and Graphics Processing Pub Date : 2022-01-07 DOI: 10.1145/3512388.3512451

Sheng-Hsiang Yu, Shengjin Wang

引用次数: 1