TY - GEN
T1 - Learning from Noisy Web Data with Category-Level Supervision
AU - Niu, Li
AU - Tang, Qingtao
AU - Veeraraghavan, Ashok
AU - Sabharwal, Ashu
N1 - Funding Information:
This work is supported by NIH 7000000356 and NSF IIS-1652633. This work is also supported by the NGA NHARP program HM0476-15-1-0007.
Publisher Copyright:
© 2018 IEEE.
PY - 2018/12/14
Y1 - 2018/12/14
N2 - Learning from web data is increasingly popular due to abundant free web resources. However, the performance gap between webly supervised learning and traditional supervised learning is still very large, due to the label noise of web data as well as the domain shift between web data and test data. To fill this gap, most existing methods propose to purify or augment web data using instance-level supervision, which generally requires heavy annotation. Instead, we propose to address the label noise and domain shift by using more accessible category-level supervision. In particular, we build our deep probabilistic framework upon variational autoencoder (VAE), in which classification network and VAE can jointly leverage category-level hybrid information. Then, we extend our method for domain adaptation followed by our low-rank refinement strategy. Extensive experiments on three benchmark datasets demonstrate the effectiveness of our proposed method.
AB - Learning from web data is increasingly popular due to abundant free web resources. However, the performance gap between webly supervised learning and traditional supervised learning is still very large, due to the label noise of web data as well as the domain shift between web data and test data. To fill this gap, most existing methods propose to purify or augment web data using instance-level supervision, which generally requires heavy annotation. Instead, we propose to address the label noise and domain shift by using more accessible category-level supervision. In particular, we build our deep probabilistic framework upon variational autoencoder (VAE), in which classification network and VAE can jointly leverage category-level hybrid information. Then, we extend our method for domain adaptation followed by our low-rank refinement strategy. Extensive experiments on three benchmark datasets demonstrate the effectiveness of our proposed method.
UR - http://www.scopus.com/inward/record.url?scp=85062839027&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85062839027&partnerID=8YFLogxK
U2 - 10.1109/CVPR.2018.00802
DO - 10.1109/CVPR.2018.00802
M3 - Conference contribution
AN - SCOPUS:85062839027
T3 - Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
SP - 7689
EP - 7698
BT - Proceedings - 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 31st Meeting of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018
Y2 - 18 June 2018 through 22 June 2018
ER -