Research on the Segmentation Method of Miao Costume Patterns Based on AD-SegNet Modeling
LUO Senyan
YANG Guiyan
CAI Jianghai
WANG Shunxia
HUANG Chengquan
WAN Linjiang
Abstract：To address the problem of diverse textures and irregular shapes of Miao costume patterns,a deep over-parameterized convolutional semantic segmentation network for attentional feature fusion(AD-SegNet)model was proposed,which added depth convolution operations to the traditional convolution operations,accelerated the convergence speed of the model,and improved the segmentation performance of the model.For the problem of large color differences in Miao costume patterns,attentional feature fusion(AFF)was introduced to enhance the extraction of feature information by fusing the feature information between the outputs of different convolutional layers,and to reduce the impact of color differences on network learning.The experimental results showed that the AD-SegNet model outperformed other comparative models and could effectively segment the Miao costume patterns.Compared with the baseline model of semantic segmentation network(SegNet),the pixel accuracy(PA),intersection over union(IoU),and dice similarity coefficient improved 0.0447、0.0921 and 0.0542 respectively,which provided an effective and feasible method for the research of ethnic minority costume pattern segmentation.
Keywords：Miao costumedeep over-parameterized convolutionAD-SegNetdepth convolution operationsAFFfeature informationSegNet
Publication Date：2024-03-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 86-91 )