Abstract
Image semantic segmentation has been widely used in medical image analysis, autonomous driving and other fields. However, the fully-supervised semantic segmentation network requires a lot of labor cost to label pixel-level training data, so weakly supervised semantic segmentation (WSSS), which requires much easily available supervision, has become a new research hotspot. This paper focuses on tackling the semantic segmentation problem under weak supervision of image-level labels. To estimate more accurate pseudo masks, this paper proposes to jointly explore sub-category clustering, context decoupling augmentation and adversarial climbing to mine more object-related regions. With sub-categories of k-means clustering, model can learn better feature representations, which breaks the dependency of object on the context by decoupling augmentation. The image is perturbed away from the classification boundary to further increase the classification score with adversarial climbing method. In order to verify the effectiveness of the method in this paper, we conduct a large number of experiments on the PASCAL VOC 2012 dataset obtained an excellent performance of 69.8% mIoU on the verification set and 69.5% mIoU on the test set, which surpassed many advanced models of the same level supervision.
This work is partially supported by National Natural Science Foundation of China (Grants no. 61772568), Guangdong Basic and Applied Basic Research Foundation (Grant no. 2019A1515012029), and Youth science and technology innovation talent of Guangdong Special Support Program.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Bearman, A., Russakovsky, O., Ferrari, V., Fei-Fei, L.: What’s the point: semantic segmentation with point supervision. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9911, pp. 549–565. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46478-7_34
Cordts, M., et al.: The cityscapes dataset for semantic urban scene understanding. In: CVPR (2016)
Zhou, B., et al.: Learning deep features for discriminative localization. In: CVPR (2016)
Su, Y., et al.: Context decoupling augmentation for weakly supervised semantic segmentation. CoRR abs/2103.01795 (2021)
Chaudhry, A., Dokania, P.K., Torr, P.: Discovering class-specific pixels for weakly-supervised semantic segmentation. In: BMVC (2017)
Wang, Y., et al.: Self-supervised equivariant attention mechanism for weakly supervised semantic segmentation. In: CVPR, pp. 12272–12281 (2020)
Lee, J., et al.: FickleNet: weakly and semi-supervised semantic image segmentation using stochastic inference. In: CVPR (2019)
Chang, Y.T., et al.: Weakly-supervised semantic segmentation via sub-category exploration. In: CVPR (2020)
Lee, J., Kim, E., Yoon, S.: Anti-adversarially manipulated attributions for weakly and semi-supervised semantic segmentation. In: CVPR (2021)
Ahn, J., Cho, S., Kwak, S.: Weakly supervised learning of instance segmentation with inter-pixel relations. In: CVPR, pp. 2209–2218 (2019)
Singh, K.K., Yong, J.L.: Hide-and-seek: forcing a network to be meticulous for weakly-supervised object and action localization. In: ICCV (2017)
Wei, Y., et al.: Object region mining with adversarial erasing: a simple classification to semantic segmentation approach. In: CVPR (2017)
Zhang, X., et al.: Adversarial complementary learning for weakly supervised object localization. CoRR abs/1804.06962 (2018)
Li, K., et al.: Tell me where to look: guided attention inference network. In: CVPR (2018)
Hou, Q., et al.: Self-erasing network for integral object attention. CoRR abs/1810.09821 (2018)
Kim, B., Kim, S.: Discriminative region suppression for weakly-supervised semantic segmentation. CoRR abs/2103.07246 (2021)
Kolesnikov, A., Lampert, C.H.: Seed, expand and constrain: three principles for weakly-supervised image segmentation. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9908, pp. 695–711. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46493-0_42
Wang, X., et al.: Weakly-supervised semantic segmentation by iteratively mining common object features. In: CVPR (2018)
Wei, Y., et al.: Revisiting dilated convolution: a simple approach for weakly- and semi-supervised semantic segmentation. In: CVPR, pp. 7268–7277 (2018)
Huang, Z., et al.: Weakly-supervised semantic segmentation network with deep seeded region growing. In: CVPR (2018)
Ma, T., Zhang, A.: AffinityNet: semi-supervised few-shot learning for disease type prediction. CoRR abs/1805.08905 (2018)
He, K., et al.: Deep residual learning for image recognition. In: CVPR, pp. 770–778 (2016)
Everingham, M., et al.: The pascal visual object classes challenge. Int. J. Comput. Vis. 88(2), 303–338 (2010)
Hariharan, B., et al.: Semantic contours from inverse detectors. In: ICCV (2011)
Chen, L.C., et al.: DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE Trans. Pattern Anal. Mach. Intell. 40(4), 834–848 (2018)
Ahn, J., Kwak, S.: Learning pixel-level semantic affinity with image-level supervision for weakly supervised semantic segmentation. In: CVPR (2018)
Li, Y., et al.: Learning dynamic routing for semantic segmentation. CoRR abs/2003.10401 (2020)
Fan, J., et al.: Learning integral objects with intra-class discriminator for weakly-supervised semantic segmentation. In: CVPR (2020)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 Springer Nature Switzerland AG
About this paper
Cite this paper
Sun, G., Yang, M., Luo, W. (2021). Adversarial Decoupling for Weakly Supervised Semantic Segmentation. In: Ma, H., et al. Pattern Recognition and Computer Vision. PRCV 2021. Lecture Notes in Computer Science(), vol 13022. Springer, Cham. https://doi.org/10.1007/978-3-030-88013-2_16
Download citation
DOI: https://doi.org/10.1007/978-3-030-88013-2_16
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-88012-5
Online ISBN: 978-3-030-88013-2
eBook Packages: Computer ScienceComputer Science (R0)