Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: On Feature Normalization and Data Augmentation
(Submitted on 25 Feb 2020 (v1), last revised 30 Mar 2021 (this version, v3))
Abstract: The moments (a.k.a., mean and standard deviation) of latent features are often removed as noise when training image recognition models, to increase stability and reduce training time. However, in the field of image generation, the moments play a much more central role. Studies have shown that the moments extracted from instance normalization and positional normalization can roughly capture style and shape information of an image. Instead of being discarded, these moments are instrumental to the generation process. In this paper we propose Moment Exchange, an implicit data augmentation method that encourages the model to utilize the moment information also for recognition models. Specifically, we replace the moments of the learned features of one training image by those of another, and also interpolate the target labels -- forcing the model to extract training signal from the moments in addition to the normalized features. As our approach is fast, operates entirely in feature space, and mixes different signals than prior methods, one can effectively combine it with existing augmentation approaches. We demonstrate its efficacy across several recognition benchmark data sets where it improves the generalization capability of highly competitive baseline networks with remarkable consistency.
Submission history
From: Boyi Li [view email][v1] Tue, 25 Feb 2020 18:59:05 GMT (652kb,D)
[v2] Fri, 6 Mar 2020 18:59:02 GMT (652kb,D)
[v3] Tue, 30 Mar 2021 18:00:00 GMT (2609kb,D)
Link back to: arXiv, form interface, contact.