《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (4): 1002-1009.DOI: 10.11772/j.issn.1001-9081.2023050623
所属专题: 第九届全国智能信息处理学术会议(NCIIP 2023)
• 第九届全国智能信息处理学术会议(NCIIP 2023) • 上一篇 下一篇
肖斌1,2, 杨模1,2, 汪敏3(), 秦光源1, 李欢4
收稿日期:
2023-05-22
修回日期:
2023-06-14
接受日期:
2023-06-15
发布日期:
2023-08-01
出版日期:
2024-04-10
通讯作者:
汪敏
作者简介:
肖斌(1978—),男,重庆人,教授,硕士,CCF会员,主要研究方向:软件工程、企业信息化基金资助:
Bin XIAO1,2, Mo YANG1,2, Min WANG3(), Guangyuan QIN1, Huan LI4
Received:
2023-05-22
Revised:
2023-06-14
Accepted:
2023-06-15
Online:
2023-08-01
Published:
2024-04-10
Contact:
Min WANG
About author:
XIAO Bin, born in 1978, M. S., professor. His research interests include software engineering, enterprise informatization.Supported by:
摘要:
针对现有的领域泛化(DG)方法对领域特征处理粗糙和泛化能力弱的问题,提出一种基于频域特征独立性这一独特视角解决领域泛化问题的方法。首先,设计频域分解算法,将图像的深度特征快速傅里叶变换(FFT)后,再从相位信息中获得领域无关特征,以提高模型对领域无关特征的识别能力;其次,基于独立性视角,通过对样本的特征赋权,进一步消除频域特征中各属性的相关性,提取最有效领域无关特征,解决样本特征之间相关性带来的泛化能力差的问题;最后,提出幅度融合策略,拉近源域和目标域的距离,进一步提升模型对未知领域的泛化能力。在流行的图像领域泛化的数据集PACS和VLCS上的实验结果表明,所提方法的准确率均值比StableNet分别高0.44、0.59个百分点,且在各个数据集上均取得了优秀的性能。
中图分类号:
肖斌, 杨模, 汪敏, 秦光源, 李欢. 独立性视角下的相频融合领域泛化方法[J]. 计算机应用, 2024, 44(4): 1002-1009.
Bin XIAO, Mo YANG, Min WANG, Guangyuan QIN, Huan LI. Domain generalization method of phase-frequency fusion from independent perspective[J]. Journal of Computer Applications, 2024, 44(4): 1002-1009.
数据集 | 类别 | 领域数 | 图像数 |
---|---|---|---|
PACS | 7 | 4 | 9 991 |
VLCS | 5 | 4 | 12 237 |
Office-Home | 65 | 4 | 15 500 |
表1 领域泛化数据集的统计数据
Tab. 1 Statistics for domain generalization datasets
数据集 | 类别 | 领域数 | 图像数 |
---|---|---|---|
PACS | 7 | 4 | 9 991 |
VLCS | 5 | 4 | 12 237 |
Office-Home | 65 | 4 | 15 500 |
方法 | PACS | 均值 | |||
---|---|---|---|---|---|
Art | Cartoon | Photo | Sketch | ||
ResNet-18 | 76.61 | 73.60 | 93.31 | 76.08 | 79.90 |
MLDG | 79.50 | 77.30 | 94.30 | 71.50 | 80.65 |
MMLD | 81.28 | 77.16 | 96.09 | 72.29 | 81.71 |
FAR | 79.30 | 77.70 | 95.30 | 74.70 | 81.75 |
MixStyle | 84.10 | 78.80 | 96.10 | 75.90 | 83.73 |
StableNet | 81.74 | 79.91 | 96.53 | 80.50 | 84.69 |
StableNet* | 83.45 | 79.60 | 94.97 | 78.92 | 84.24 |
RSC | 83.43 | 80.31 | 95.99 | 80.85 | 85.15 |
RSC* | 80.55 | 78.60 | 94.43 | 76.02 | 82.40 |
MASF | 80.29 | 77.17 | 94.99 | 71.79 | 81.06 |
FACT | 85.37 | 78.38 | 95.15 | 79.15 | 84.51 |
本文方法 | 84.47 | 81.27 | 94.61 | 80.17 | 85.13 |
表2 各方法在PACS数据集上的Top-1分类准确率比较 (%)
Tab. 2 Comparison of Top-1 classification accuracy of different methods on PACS dataset
方法 | PACS | 均值 | |||
---|---|---|---|---|---|
Art | Cartoon | Photo | Sketch | ||
ResNet-18 | 76.61 | 73.60 | 93.31 | 76.08 | 79.90 |
MLDG | 79.50 | 77.30 | 94.30 | 71.50 | 80.65 |
MMLD | 81.28 | 77.16 | 96.09 | 72.29 | 81.71 |
FAR | 79.30 | 77.70 | 95.30 | 74.70 | 81.75 |
MixStyle | 84.10 | 78.80 | 96.10 | 75.90 | 83.73 |
StableNet | 81.74 | 79.91 | 96.53 | 80.50 | 84.69 |
StableNet* | 83.45 | 79.60 | 94.97 | 78.92 | 84.24 |
RSC | 83.43 | 80.31 | 95.99 | 80.85 | 85.15 |
RSC* | 80.55 | 78.60 | 94.43 | 76.02 | 82.40 |
MASF | 80.29 | 77.17 | 94.99 | 71.79 | 81.06 |
FACT | 85.37 | 78.38 | 95.15 | 79.15 | 84.51 |
本文方法 | 84.47 | 81.27 | 94.61 | 80.17 | 85.13 |
算法 | Office-Home | 均值 | |||
---|---|---|---|---|---|
Art | Clipart | Product | Real-World | ||
MMD-AAE | 56.50 | 47.30 | 72.10 | 74.80 | 62.68 |
DADG | 55.57 | 48.71 | 70.90 | 73.70 | 62.22 |
JiGen | 53.04 | 47.51 | 71.47 | 72.79 | 61.20 |
RSC | 58.42 | 47.90 | 71.63 | 74.54 | 63.12 |
SagNet | 60.20 | 45.30 | 70.40 | 73.30 | 62.30 |
本文方法 | 57.10 | 52.42 | 72.25 | 73.90 | 63.92 |
表3 各方法在Office?Home数据集上的Top-1分类准确率比较 (%)
Tab. 3 Comparison of Top-1 classification accuracy of different methods on Office-Home dataset
算法 | Office-Home | 均值 | |||
---|---|---|---|---|---|
Art | Clipart | Product | Real-World | ||
MMD-AAE | 56.50 | 47.30 | 72.10 | 74.80 | 62.68 |
DADG | 55.57 | 48.71 | 70.90 | 73.70 | 62.22 |
JiGen | 53.04 | 47.51 | 71.47 | 72.79 | 61.20 |
RSC | 58.42 | 47.90 | 71.63 | 74.54 | 63.12 |
SagNet | 60.20 | 45.30 | 70.40 | 73.30 | 62.30 |
本文方法 | 57.10 | 52.42 | 72.25 | 73.90 | 63.92 |
算法 | VLCS | 均值 | |||
---|---|---|---|---|---|
Caltech | LabelMe | VOC | Sun | ||
JiGen | 96.17 | 62.06 | 70.93 | 71.40 | 75.14 |
M-ADA | 74.33 | 48.38 | 45.13 | 33.82 | 50.42 |
MMLD | 97.01 | 62.20 | 73.01 | 72.49 | 76.18 |
RSC | 96.21 | 62.51 | 73.81 | 72.10 | 76.16 |
StableNet | 96.67 | 65.36 | 73.59 | 74.97 | 77.65 |
本文方法 | 96.75 | 67.17 | 74.61 | 74.41 | 78.24 |
表4 各方法在VLCS数据集上的Top-1分类准确率比较 (%)
Tab. 4 Comparison of Top-1 classification accuracy of different methods on VLCS dataset
算法 | VLCS | 均值 | |||
---|---|---|---|---|---|
Caltech | LabelMe | VOC | Sun | ||
JiGen | 96.17 | 62.06 | 70.93 | 71.40 | 75.14 |
M-ADA | 74.33 | 48.38 | 45.13 | 33.82 | 50.42 |
MMLD | 97.01 | 62.20 | 73.01 | 72.49 | 76.18 |
RSC | 96.21 | 62.51 | 73.81 | 72.10 | 76.16 |
StableNet | 96.67 | 65.36 | 73.59 | 74.97 | 77.65 |
本文方法 | 96.75 | 67.17 | 74.61 | 74.41 | 78.24 |
幅度融合 | 特征独立性策略 | Art | Cartoo | Photo | Sketch | 均值 |
---|---|---|---|---|---|---|
— | — | 76.61 | 73.60 | 93.31 | 76.08 | 79.90 |
√ | — | 83.31 | 81.23 | 95.63 | 77.45 | 84.41 |
— | √ | 83.44 | 79.60 | 94.97 | 78.92 | 84.32 |
√ | √ | 84.47 | 81.27 | 94.61 | 80.17 | 85.13 |
表5 PACS数据集上的消融实验结果 (%)
Tab. 5 Ablation experiment results on PACS datasets
幅度融合 | 特征独立性策略 | Art | Cartoo | Photo | Sketch | 均值 |
---|---|---|---|---|---|---|
— | — | 76.61 | 73.60 | 93.31 | 76.08 | 79.90 |
√ | — | 83.31 | 81.23 | 95.63 | 77.45 | 84.41 |
— | √ | 83.44 | 79.60 | 94.97 | 78.92 | 84.32 |
√ | √ | 84.47 | 81.27 | 94.61 | 80.17 | 85.13 |
1 | KARPATHY A, TODERICI G, SHETTY S, et al. Large-scale video classification with convolutional neural networks [C]// Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2014: 1725-1732. 10.1109/cvpr.2014.223 |
2 | 范苍宁,刘鹏,肖婷,等.深度域适应综述:一般情况与复杂情况[J].自动化学报,2021,47(3):515-548. 10.16383/j.aas.c200238 |
FAN C N, LIU P, XIAO T, et al. A review on depth domain adaption: general situation and complex situation[J]. Acta Automatica Sinica, 2021,47(3): 515-548. 10.16383/j.aas.c200238 | |
3 | TAN C, SUN F, KONG T, et al. A survey on deep transfer learning [C]// Proceedings of the 2018 International Conference on Artificial Neural Networks: Artificial Neural Networks and Machine Learning. Cham: Springer, 2018: 270-279. 10.1007/978-3-030-01424-7_27 |
4 | WANG J, LAN C, LIU C, et al. Generalizing to unseen domains: a survey on domain generalization [C]// Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence.[S.l.]: IJCAI, 2021: 4627-4635. 10.24963/ijcai.2021/628 |
5 | PRAKASH A, BOOCHOON S, BROPHY M, et al. Structured domain randomization: bridging the reality gap by context-aware synthetic data [C]// Proceedings of the 2019 International Conference on Robotics and Automation. Piscataway: IEEE, 2019: 7249-7255. 10.1109/icra.2019.8794443 |
6 | HUANG J, GUAN D, XIAO A, et al. FSDR: frequency space domain randomization for domain generalization [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 6887-6898. 10.1109/cvpr46437.2021.00682 |
7 | ZHOU K Y, YANG Y X, HOSPEDALES T, et al. Deep domain-adversarial image generation for domain generalisation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 13025-13032. 10.1609/aaai.v34i07.7003 |
8 | ARJOVSKY M, BOTTOU L, GULRAJANI I, et al. Invariant risk minimization[EB/OL]. (2019-07-05)[2023-05-01]. . |
9 | ZHANG H, ZHANG Y-F, LIU W, et al. Towards principled disentanglement for domain generalization [C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 8014-8024. 10.1109/cvpr52688.2022.00786 |
10 | LIN Y, DONG H, WANG H, et al. Bayesian invariant risk minimization [C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 16000-16009. 10.1109/cvpr52688.2022.01555 |
11 | KIM D, YOO Y, PARK S, et al. SelfReg: self-supervised contrastive regularization for domain generalization [C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 9599-9608. 10.1109/iccv48922.2021.00948 |
12 | XU J, XIAO L, LÓPEZ A M. Self-supervised domain adaptation for computer vision tasks[J]. IEEE Access, 2019, 7: 156694-156706. 10.1109/access.2019.2949697 |
13 | VENKATESWARA H, EUSEBIO J, CHAKRABORTY S, et al. Deep hashing network for unsupervised domain adaptation[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 5385-5394. 10.1109/cvpr.2017.572 |
14 | WEN J, LIU R, ZHENG N, et al. Exploiting local feature patterns for unsupervised domain adaptation[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 5401-5408. 10.1609/aaai.v33i01.33015401 |
15 | PAN S J, TSANG I W, KWOK J T, et al. Domain adaptation via transfer component analysis[J]. IEEE Transactions on Neural Networks, 2011, 22(2): 199-210. 10.1109/tnn.2010.2091281 |
16 | TZENG E, HOFFMAN J, ZHANG N, et al. Deep domain confusion: maximizing for domain invariance [EB/OL]. (2014-12-10) [2023-05-01]. . |
17 | KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90. 10.1145/3065386 |
18 | ZHU Y, ZHUANG F, WANG J, et al. Deep subdomain adaptation network for image classification[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(4): 1713-1722. 10.1109/tnnls.2020.2988928 |
19 | SHU Y, CAO Z, WANG C, et al. Open domain generalization with domain-augmented meta-learning [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 9619-9628. 10.1109/cvpr46437.2021.00950 |
20 | XU Q, ZHANG R, ZHANG Y, et al. A Fourier-based framework for domain generalization [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, DC: IEEE Computer Society, 2021: 14378-14387. 10.1109/cvpr46437.2021.01415 |
21 | HU S, ZHANG K, CHEN Z, et al. Domain generalization via multidomain discriminant analysis [C]// Proceedings of the 35th Uncertainty in Artificial Intelligence Conference. New York: JMLR.org, 2020: 292-302. |
22 | MITROVIC J, McWILLIAMS B, WALKER J, et al. Representation learning via invariant causal mechanisms [EB/OL]. (2020-10-15)[2023-05-01]. . |
23 | SHI Y, SEELY J, TORR P H S, et al. Gradient matching for domain generalization[EB/OL]. (2021-04-20)[2023-05-01]. . |
24 | DUBEY A, RAMANATHAN V, PENTLAND A, et al. Adaptive methods for real-world domain generalization [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE,2021: 14335-14344. 10.1109/cvpr46437.2021.01411 |
25 | RAHIMI A, RECHT B. Random features for large-scale kernel machines [C]// Proceedings of the 20th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2007: 1177-1184. 10.7551/mitpress/7503.003.0144 |
26 | LE Q V, SARLOS T, SMOLA A J. Fastfood: approximate kernel expansions in loglinear time [EB/OL]. (2014-08-13)[2023-05-01]. . |
27 | ATARASHI K, MAJI S, OYAMA S. Random feature maps for the itemset kernel[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 3199-3206. 10.1609/aaai.v33i01.33013199 |
28 | HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 770-778. 10.1109/cvpr.2016.90 |
29 | LI D, YANG Y, SONG Y-Z, et al. Learning to generalize: meta-learning for domain generalization[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2018, 32(1): 3490-3497. 10.1609/aaai.v32i1.11596 |
30 | MATSUURA T, HARADA T. Domain generalization using a mixture of multiple latent domains[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 11749-11756. 10.1609/aaai.v34i07.6846 |
31 | JIN X, LAN C, ZENG W, et al. Feature alignment and restoration for domain generalization and adaptation [EB/OL]. (2020-06-22) [2023-05-01]. . 10.1109/tmm.2021.3104379 |
32 | HUANG Z, WANG H, XING E P, et al. Self-challenging improves cross-domain generalization [C]// Proceedings of the 2020 European Conference on Computer Vision. Cham: Springer, 2020: 124-140. 10.1007/978-3-030-58536-5_8 |
33 | ZHANG X, CUI P, XU R, et al. Deep stable learning for out-of-distribution generalization [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 5368-5378. 10.1109/cvpr46437.2021.00533 |
34 | ZHOU K, YANG Y, QIAO Y, et al. Domain generalization with MixStyle [EB/OL]. (2021-04-05)[2023-05-01]. . 10.1007/s11263-023-01913-8 |
35 | DOU Q, CASTRO D C, KAMNITSAS K, et al. Domain generalization via model-agnostic learning of semantic features [C]// Proceedings of the 33rd International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2019: 6450‑6461. |
36 | LI H, PAN S J, WANG S, et al. Domain generalization with adversarial feature learning [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE,2018: 5400-5409. 10.1109/cvpr.2018.00566 |
37 | CHEN K, ZHUANG D, CHANG J M. Discriminative adversarial domain generalization with meta-learning based cross-domain validation[J]. Neurocomputing, 2022, 467: 418-426. 10.1016/j.neucom.2021.09.046 |
38 | CARLUCCI F M, D’INNOCENTE A, BUCCI S, et al. Domain generalization by solving jigsaw puzzles [C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 2224-2233. 10.1109/cvpr.2019.00233 |
39 | NAM H, LEE H, PARK J, et al. Reducing domain gap by reducing style bias [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE, 2021: 8686-8695. 10.1109/cvpr46437.2021.00858 |
40 | QIAO F, ZHAO L, PENG X. Learning to learn single domain generalization [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 12553-12562. 10.1109/cvpr42600.2020.01257 |
[1] | 石锐, 李勇, 朱延晗. 基于特征梯度均值化的调制信号对抗样本攻击算法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2521-2527. |
[2] | 王东炜, 刘柏辰, 韩志, 王艳美, 唐延东. 基于低秩分解和向量量化的深度网络压缩方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 1987-1994. |
[3] | 王美, 苏雪松, 刘佳, 殷若南, 黄珊. 时频域多尺度交叉注意力融合的时间序列分类方法[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1842-1847. |
[4] | 翟飞宇, 马汉达. 基于DenseNet的经典-量子混合分类模型[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1905-1910. |
[5] | 颜梦玫, 杨冬平. 深度神经网络平均场理论综述[J]. 《计算机应用》唯一官方网站, 2024, 44(2): 331-343. |
[6] | 柴汶泽, 范菁, 孙书魁, 梁一鸣, 刘竟锋. 深度度量学习综述[J]. 《计算机应用》唯一官方网站, 2024, 44(10): 2995-3010. |
[7] | 谢莉, 舒卫平, 耿俊杰, 王琼, 杨海麟. 结合加权原型和自适应张量子空间的小样本宫颈细胞分类[J]. 《计算机应用》唯一官方网站, 2024, 44(10): 3200-3208. |
[8] | 周雯, 谌雨章, 温志远, 王诗琦. 基于位置编码重叠切块嵌入和多尺度通道交互注意力的鱼类图像分类[J]. 《计算机应用》唯一官方网站, 2024, 44(10): 3209-3216. |
[9] | 陈彤, 位纪伟, 何仕远, 宋井宽, 杨阳. 基于自适应攻击强度的对抗训练方法[J]. 《计算机应用》唯一官方网站, 2024, 44(1): 94-100. |
[10] | 赵旭剑, 李杭霖. 基于混合机制的深度神经网络压缩算法[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2686-2691. |
[11] | 申云飞, 申飞, 李芳, 张俊. 基于张量虚拟机的深度神经网络模型加速方法[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2836-2844. |
[12] | 李校林, 杨松佳. 基于深度学习的多用户毫米波中继网络混合波束赋形[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2511-2516. |
[13] | 李淦, 牛洺第, 陈路, 杨静, 闫涛, 陈斌. 融合视觉特征增强机制的机器人弱光环境抓取检测[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2564-2571. |
[14] | 王彬, 向甜, 吕艺东, 王晓帆. 基于NSGA‑Ⅱ的自适应多尺度特征通道分组优化算法[J]. 《计算机应用》唯一官方网站, 2023, 43(5): 1401-1408. |
[15] | 杨海宇, 郭文普, 康凯. 基于卷积长短时深度神经网络的信号调制方式识别方法[J]. 《计算机应用》唯一官方网站, 2023, 43(4): 1318-1322. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||