Abstract
Unsupervised Domain Adaptation (UDA) challenges the problem of alleviating the effect of domain shift. Common UDA methods all require labelled source samples. However, in some real application scenarios, such as Federated Learning, the source data is inaccessible due to data privacy or intellectual property, and only a pre-trained source model and target data without labels are accessible. This challenging problem is called Source-Free Domain Adaptation (SFDA). To address this, we introduce a generation encoder to generate source prototypes depending on the hidden knowledge from the pre-trained source classifier. The generated source prototypes can describe the distribution of source samples in the feature space to a certain extent to solve the Source-Free problem. We also propose Source-Free Implicit Semantic Augmentation (SFISA) for adaptation. SFISA consists of two main stages: source and target class prototypes generation and Source-Free semantic augmentation adaptation based on generated class prototypes. Extensive experiments on the UDA benchmarks demonstrate the efficacy of our generation encoder and augmentation method SFISA.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Ben-David, S., Blitzer, J., Crammer, K., Kulesza, A., Pereira, F., Vaughan, J.W.: A theory of learning from different domains. Mach. Learn. 79(1), 151–175 (2009). https://doi.org/10.1007/s10994-009-5152-4
Choi, J., Kim, T., Kim, C.: Self-ensembling with GAN-based data augmentation for domain adaptation in semantic segmentation. In: ICCV, pp. 6829–6839 (2019)
Cui, S., Wang, S., Zhuo, J., Li, L., Huang, Q., Tian, Q.: Towards discriminability and diversity: batch nuclear-norm maximization under label insufficient situations. In: CVPR, pp. 3940–3949 (2020)
Cui, S., Wang, S., Zhuo, J., Li, L., Huang, Q., Tian, Q.: Fast batch nuclear-norm maximization and minimization for robust domain adaptation. CoRR (2021)
Cui, S., Wang, S., Zhuo, J., Su, C., Huang, Q., Tian, Q.: Gradually vanishing bridge for adversarial domain adaptation. In: CVPR, pp. 12452–12461 (2020)
Dai, S., Cheng, Y., Zhang, Y., Gan, Z., Liu, J., Carin, L.: Contrastively smoothed class alignment for unsupervised domain adaptation. In: ACCV, pp. 268–283 (2020)
Ganin, Y., Lempitsky, V.S.: Unsupervised domain adaptation by backpropagation. In: ICML, pp. 1180–1189 (2015)
Gu, J., Dong, C.: Interpreting super-resolution networks with local attribution maps. In: CVPR, pp. 9199–9208 (2021)
He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR, pp. 770–778 (2016)
Huang, J., Guan, D., Xiao, A., Lu, S.: Model adaptation: historical contrastive learning for unsupervised domain adaptation without source data. In: NeurIPS, pp. 3635–3649 (2021)
Jing, L., Vincent, P., LeCun, Y., Tian, Y.: Understanding dimensional collapse in contrastive self-supervised learning. CoRR (2021)
Kang, G., Jiang, L., Yang, Y., Hauptmann, A.G.: Contrastive adaptation network for unsupervised domain adaptation. In: CVPR, pp. 4893–4902 (2019)
Lee, C., Batra, T., Baig, M.H., Ulbricht, D.: Sliced wasserstein discrepancy for unsupervised domain adaptation. In: CVPR, pp. 10285–10295 (2019)
Li, R., Jiao, Q., Cao, W., Wong, H., Wu, S.: Model adaptation: unsupervised domain adaptation without source data. In: CVPR, pp. 9638–9647 (2020)
Li, S., Xie, M., Gong, K., Liu, C.H., Wang, Y., Li, W.: Transferable semantic augmentation for domain adaptation. In: CVPR, pp. 11516–11525 (2021)
Li, X., et al.: PointFlow: flowing semantics through points for aerial image segmentation. In: CVPR, pp. 4217–4226 (2021)
Liang, J., Hu, D., Feng, J.: Do we really need to access the source data? Source hypothesis transfer for unsupervised domain adaptation. In: ICML, pp. 6028–6039 (2020)
Liang, J., Hu, D., Feng, J.: Domain adaptation with auxiliary target domain-oriented classifier. In: CVPR, pp. 16632–16642 (2021)
Liu, H., Long, M., Wang, J., Jordan, M.I.: Transferable adversarial training: a general approach to adapting deep classifiers. In: ICML, pp. 4013–4022 (2019)
Long, M., Cao, Y., Wang, J., Jordan, M.I.: Learning transferable features with deep adaptation networks. In: ICML, pp. 97–105 (2015)
Long, M., Zhu, H., Wang, J., Jordan, M.I.: Deep transfer learning with joint adaptation networks. In: ICML, pp. 2208–2217 (2017)
McShane, E.J.: Jensen’s inequality. BAMS, pp. 521–527 (1937)
Mei, H., Wan, T., Eisner, J.: Noise-contrastive estimation for multivariate point processes. In: NeurIPS, pp. 5204–5214 (2020)
Müller, R., Kornblith, S., Hinton, G.E.: When does label smoothing help? In: NeurIPS, pp. 4696–4705 (2019)
Pei, Z., Cao, Z., Long, M., Wang, J.: Multi-adversarial domain adaptation. In: AAAI, pp. 3934–3941 (2018)
Qiu, Z., et al.: Source-free domain adaptation via avatar prototype generation and adaptation. In: IJCAI, pp. 2921–2927 (2021)
Saenko, K., Kulis, B., Fritz, M., Darrell, T.: Adapting visual category models to new domains. In: ECCV, pp. 213–226 (2010)
Venkateswara, H., Eusebio, J., Chakraborty, S., Panchanathan, S.: Deep hashing network for unsupervised domain adaptation. In: CVPR, pp. 5385–5394 (2017)
Xu, S., et al.: Generative low-bitwidth data free quantization. In: ECCV, pp. 1–17 (2020)
Yang, S., Wang, Y., van de Weijer, J., Herranz, L., Jui, S.: Unsupervised domain adaptation without source data by casting a BAIT. CoRR (2020)
Yang, S., Wang, Y., van de Weijer, J., Herranz, L., Jui, S.: Exploiting the intrinsic neighborhood structure for source-free domain adaptation. In: NeurIPS, pp. 29393–29405 (2021)
Zhang, Y., Tang, H., Jia, K., Tan, M.: Domain-symmetric networks for adversarial domain adaptation. In: CVPR, pp. 5031–5040 (2019)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Zhang, Z., Zhang, Z. (2022). Source-Free Implicit Semantic Augmentation for Domain Adaptation. In: Khanna, S., Cao, J., Bai, Q., Xu, G. (eds) PRICAI 2022: Trends in Artificial Intelligence. PRICAI 2022. Lecture Notes in Computer Science, vol 13630. Springer, Cham. https://doi.org/10.1007/978-3-031-20865-2_2
Download citation
DOI: https://doi.org/10.1007/978-3-031-20865-2_2
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-20864-5
Online ISBN: 978-3-031-20865-2
eBook Packages: Computer ScienceComputer Science (R0)