ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

Improving GANs for Long-Tailed Data Through Group Spectral Regularization

Rangwani, H and Jaswani, N and Karmali, T and Jampani, V and Babu, RV (2022) Improving GANs for Long-Tailed Data Through Group Spectral Regularization. In: 17th European Conference on Computer Vision, ECCV 2022, 23 - 27 October 2022, Tel Aviv, pp. 426-442.

[img]
Preview
PDF
ECCV_2022.pdf - Published Version

Download (11MB) | Preview
Official URL: https://doi.org/10.1007/978-3-031-19784-0_25

Abstract

Deep long-tailed learning aims to train useful deep networks on practical, real-world imbalanced distributions, wherein most labels of the tail classes are associated with a few samples. There has been a large body of work to train discriminative models for visual recognition on long-tailed distribution. In contrast, we aim to train conditional Generative Adversarial Networks, a class of image generation models on long-tailed distributions. We find that similar to recognition, state-of-the-art methods for image generation also suffer from performance degradation on tail classes. The performance degradation is mainly due to class-specific mode collapse for tail classes, which we observe to be correlated with the spectral explosion of the conditioning parameter matrix. We propose a novel group Spectral Regularizer (gSR) that prevents the spectral explosion alleviating mode collapse, which results in diverse and plausible image generation even for tail classes. We find that gSR effectively combines with existing augmentation and regularization techniques, leading to state-of-the-art image generation performance on long-tailed data. Extensive experiments demonstrate the efficacy of our regularizer on long-tailed datasets with different degrees of imbalance. Project Page: https://sites.google.com/view/gsr-eccv22. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Item Type: Conference Paper
Publication: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher: Springer Science and Business Media Deutschland GmbH
Additional Information: The copyright for this article belongs to Author(S).
Keywords: Deep learning, Discriminative models; Image generations; Long-tailed distributions; Parameter matrices; Performance degradation; Real-world; Regularisation; Regularizer; State-of-the-art methods; Visual recognition, Generative adversarial networks
Department/Centre: Division of Interdisciplinary Sciences > Computational and Data Sciences
Date Deposited: 09 Jan 2023 08:49
Last Modified: 09 Jan 2023 08:49
URI: https://eprints.iisc.ac.in/id/eprint/78925

Actions (login required)

View Item View Item