计算机科学 ›› 2024, Vol. 51 ›› Issue (11A): 231200162-8.doi: 10.11896/jsjkx.231200162
王信超, 余映, 陈安, 赵辉荣
WANG Xinchao, YU Ying, CHEN An, ZHAO Huirong
摘要： 针对古代壁画脱落区域难以准确自动标定的问题,文中提出了一种基于多尺度融合注意力网络的轻量化网络模型。首先,提出多尺度融合注意力模块使网络能够学习到更多不同尺度的特征,并重点关注其中更重要的特征,从而提高标定壁画脱落区域的准确率。在提出的多尺度融合注意力模块中采用了深度可分离卷积,使网络模型更加轻量化。其次,采用交叉熵损失与Dice得分相结合的方式作为损失函数,并采用Adam优化器进一步提高标定壁画脱落区域的准确率。此外,构建了敦煌莫高窟壁画和云南石屏罗色庙壁画数据集,并对其脱落区域进行了人工标定。实验结果表明,所提网络模型能够准确地标定出古代壁画中的脱落病害区域。与现有深度学习方法进行对比,所提模型的参数量显著减少,且在主观视觉质量、客观评价指标以及泛化性能上都有更好的表现。
中图分类号:
| [1]ROSA DE A,BONACEHI A M,CAPPELLINI V,et al.Image segmentation and region filling for virtual restoration of artworks [C]//Proceedings 2001 International Conference on Image Processing.Thessaloniki:IEEE,2001:562-565. [2]LIU J M.Research on the protection and intelligent restoration technology of ancient mural images [D].Hangzhou:Zhejiang University,2010. [3]TURAKHIA N,SHAH R,JOSHI M.Automatic crack detec-tion in heritage site images for image inpainting [C]//Proceedings of the Eighth Indian Conference on Computer Vision,Graphics and Image Processing.Mumbai:ACM,2012:1-8. [4]WU M,WANG H Q,LI W Y.Research on multi-scale detection and image inpainting of Tang dynastytomb murals [J].Computer Engineering and Applications,2016,52(11):169-174. [5]LI C Y,WANG H Q,WU M,et al.Automatic recognition and virtual restoration of mud spot disease of Tang dynasty tomb murals image [J].Computer Engineering and Applications,2016,52(15):233-236. [6]CAO J F,LI Y F,CUI H Y,et al.The Application of Improved Region Growing Algorithm for the Automatic Calibration of Shedding Disease [J].Journal of Xinjiang University(Natural Science Edition),2018,35(4):429-436. [7]DENG X,YU Y.Automatic calibration of crack and flaking dis-eases in ancient temple murals [J].Heritage Science,2022,10(1):163. [8]HUANG R,FENG W,FAN M,et al.Learning multi-path CNNfor mural deterioration detection [J].Journal of Ambient Intelligence and Humanized Computing,2020,11:3101-3108. [9]LIN Y,XU C,LYU S.Disease Regions Recognition on MuralHyperspectral Images Combined by MNF and BP Neural Network [C]//Journal of Physics Conference Series.Qingdao:IOP,2019:104-110. [10]LÜ S Q,WANG S H,HOU M L,et al.Extraction of mural paint loss diseases based on improved U-Net [J].Geomatics World,2022,29(1):69-74. [11]SHELHAMER E,LONG J,DARRELL T.Fully convolutional networks for semantic segmentation [J].IEEE Trans Pattern Anal Mach Intell,2017,39(4):640-651. [12]RONNEBERGER O,FISCHER P,BROX T.U-Net:convolu-tional networks for biomedical image segmentation [C]//Proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Inter-vention.Munich:Springer,2015:234-241. [13]ALOM M Z,HASAN M,YAKOPCIC C,et al.Recurrent residual convolutional neural network based on u-net(r2u-net)for medical image segmentation [EB/OL].(2018-02-20)[2018-05-29].https://arxiv.org/abs/1802.06955. [14]OKTAY O,SCHLEMPER J,FOLGOC L L,et al.Attention u-net:Learning where to look for the pancreas [EB/OL].(2018-04-11)[2018-05-20].https://arxiv.org/abs/1804.03999. [15]ZHOU Z,RAHMAN SIDDIQUEE M M,TAJBAKHSH N,et al.Unet++:A nested u-net architecture for medical image segmentation [C]//Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support Cham.Springer,2018:3-11. [16]CHEN J,LU Y,YU Q,et al.Transunet:Transformers makestrong encoders for medical image segmentation [EB/OL].(2021-02-18)[2021-02-18].https://arxiv.org/abs/2102.04306. [17]CAO H,WANG Y,CHEN J,et al.Swin-unet:Unet-like puretransformer for medical image segmentation [C]//European Conference on Computer Vision.Cham:Springer,2022:205-218. [18]KIM N,KIM D,LAN C,et al.Restr:Convolution-free referringimage segmentation using transformers [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:18145-18154. [19]RU L,ZHAN Y,YU B,et al.Learning affinity from attention:End-to-end weakly-supervised semantic segmentation with transformers [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:16846-16855. [20]ZHANG J,YANG K,MA C,et al.Bending reality:Distortion-aware transformers for adapting to panoramic semantic segmentation [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:16917-16927. [21]ZHANG H,LI F,XU H,et al.MP-Former:Mask-piloted transformer for image segmentation [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Vancouver:IEEE,2023:18074-18083. [22]ZHANG C,BENGIO S,HARDT M,et al.Understanding deep learning(still)requires rethinking generalization[J].Communications of the ACM,2021,64(3):107-115. [23]WOO S,PARK J,LEE J Y,et al.Cbam:Convolutional block attention module [C]//Proceedings of the European Conference on Computer Vision.Cham:Springer,2018:3-19. [24]CHOLLET F.Xception:Deep learning with depthwise separable convolutions [C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Honolulu:IEEE,2017:1251-1258. [25]BADRINARAYANAN V,KENDALL A,CIPOLLA R.Segnet:A deep convolutional encoder-decoder architecture for image segmentation [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(12):2481-2495. [26]CHEN G P,LI L,DAI Y,et al.NU-net:An unpretentious nested U-net for breast tumor segmentation [EB/OL].(2022-09-15)[2022-12-13].https://arxiv.org/abs/2209.07193 [27]PASZKE A,CHAURASIA A,KIM S,et al.Enet:A deep neural network architecture for real-time semantic segmentation[J].arXiv:1606.02147,2016. [28]DINH B D,NGUYEN T T,TRAN T T,et al.1M parameters are enough? A lightweight CNN-based model for medical image segmentation[C]//2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference(APSIPA ASC).IEEE,2023:1279-1284. |
|
||
首页