[1]曾浩文,汪慧兰*,赵 侃,等.基于 SOLOV2 改进的实例分割算法研究[J].计算机技术与发展,2023,33(09):45-51.[doi:10. 3969 / j. issn. 1673-629X. 2023. 09. 007]
 ZENG Hao-wen,WANG Hui-lan*,ZHAO Kan,et al.Research on Improved Instance Segmentation Algorithm Based on SOLOV2[J].,2023,33(09):45-51.[doi:10. 3969 / j. issn. 1673-629X. 2023. 09. 007]
点击复制

基于 SOLOV2 改进的实例分割算法研究()
分享到:

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
33
期数:
2023年09期
页码:
45-51
栏目:
媒体计算
出版日期:
2023-09-10

文章信息/Info

Title:
Research on Improved Instance Segmentation Algorithm Based on SOLOV2
文章编号:
1673-629X(2023)09-0045-07
作者:
曾浩文汪慧兰* 赵 侃王桂丽
安徽师范大学 物理与电子信息学院,安徽 芜湖 241002
Author(s):
ZENG Hao-wenWANG Hui-lan* ZHAO KanWANG Gui-li
School of Physics and Electronic Information,Anhui Normal University,Wuhu 241002,China
关键词:
实例分割ResNextSOLOV2特征金子塔网络NAS-FPN
Keywords:
instance segmentationResNextSOLOV2feature pyramid networkNAS-FPN
分类号:
TP301
DOI:
10. 3969 / j. issn. 1673-629X. 2023. 09. 007
摘要:
实例分割在图像分类的基础上为每一个物体生成像素级别的分割掩码,是当前计算机视觉领域热门研究课题,也是极具挑战性的任务之一。 针对当前算法存在的分割精度和鲁棒性不高等问题,提出了一种改进的 SOLOV2 算法。 首先,以 FCN( Fully Convolutional Networks for Semantic Segmentation)算法为整体框架,采用 ResNext 作为骨干网络,在不增加网络参数量和计算量的前提下可以有效提升网络的精度; 其次, 采用改进的 NAS - FPN ( Neural Architecture Search FeaturePyramid Network) 作为特征金字塔网络结构,这是一种可以在 FPN 中进行特征图的搜索和组合结构,使网络可以重新搜索并融合已经提取的特征图,以此来解决网络不能充分感知特征图从而导致网络精度不高的问题;最后,通过调整超参数得到整个分割网络模型。 通过在 COCO2017 数据集上与 BDD100K 数据集上进行实验分析比较可知,改进的基于 SOLOV2 实例分割算法精度达到 41. 8% ,在兼顾实时性的同时网络精度提升了 2. 1% 。 通过实验证明改进的算法可以适应多种交通场景,可以完成交通场景目标的检测与分割。
Abstract:
Instance segmentation generates pixel - level segmentation masks for each object based on image classification, which iscurrently one of the popular research?
topics and challenging tasks in computer vision. To address the problems of poor segmentationaccuracy and robustness of current algorithms,we propose an?
improved SOLOV2 algorithm. Firstly,FCN ( Fully Convolutional Networksfor Semantic Segmentation) is used as the overall framework,and ResNext is adopted?
as the backbone network,which can effectivelyimprove the accuracy of the network without raising the number of network parameters and computational effort. Secondly,a modifiedNAS-FPN ( Neural Architecture Search Feature Pyramid Network) is used as the feature pyramid network structure,which is a structurethat?
allows the search and combination of feature maps in the FPN,so that the network can re - search and fuse the already extractedfeature maps,as a solution to the problem that the network cannot fully perceive the feature maps and thus the network accuracy is nothigh. Finally, the whole segmentation network model is obtained by adjusting the hyperparameters. The experimental analysis andcomparison on the COCO2017 dataset and the BDD100K dataset shows that the improved SOLOV2 instance segmentation algorithmachieves 41. 8% accuracy,which improves the network accuracy by 2. 1% while taking into account the real - time performance. It isproved through experiments that the improved algorithm can adapt to a variety of traffic scenes and can complete the detection and segmentation of traffic scene targets.

相似文献/References:

[1]姜世浩,齐苏敏,王来花,等.基于 Mask R-CNN 和多特征融合的实例分割[J].计算机技术与发展,2020,30(09):65.[doi:10. 3969 / j. issn. 1673-629X. 2020. 09. 012]
 JIANG Shi-hao,QI Su-min,WANG Lai-hua,et al.Instance Segmentation Modal Based on Mask R-CNN and Multi-feature Fusion[J].,2020,30(09):65.[doi:10. 3969 / j. issn. 1673-629X. 2020. 09. 012]
[2]张婧婧.基于深度学习和图像处理的马体尺测量设计[J].计算机技术与发展,2020,30(11):180.[doi:10. 3969 / j. issn. 1673-629X. 2020. 11. 033]
 ZHANG Jing-jing.Measurement and Design of Horse爷s Body Size Based on Deep Learning and Image Processing Technology[J].,2020,30(09):180.[doi:10. 3969 / j. issn. 1673-629X. 2020. 11. 033]

更新日期/Last Update: 2023-09-10