Computer Science > Computer Vision and Pattern Recognition
[Submitted on 5 Aug 2021 (this version), latest version 13 Sep 2021 (v4)]
Title:dp-GAN : Alleviating Mode Collapse in GAN via Diversity Penalty Module
View PDFAbstract:The vanilla GAN [5] suffers from mode collapse deeply, which usually manifests as that the images generated by generators tend to have a high similarity amongst them, even though their corresponding latent vectors have been very different. In this paper, we introduce a pluggable block called diversity penalty (dp) to alleviate mode collapse of GANs. It is used to reduce the similarity of image pairs in feature space, i.e., if two latent vectors are different, then we enforce the generator to generate two images with different features. The normalized Gram Matrix is used to measure the similarity. We compare the proposed method with Unrolled GAN [17], BourGAN [26], PacGAN [14], VEEGAN [23] and ALI [4] on 2D synthetic dataset, and results show that our proposed method can help GAN capture more modes of the data distribution. Further, we apply this penalty term into image data augmentation on MNIST, Fashion-MNIST and CIFAR-10, and the testing accuracy is improved by 0.24%, 1.34% and 0.52% compared with WGAN GP [6], respectively. Finally, we quantitatively evaluate the proposed method with IS and FID on CelebA, CIFAR-10, MNIST and Fashion-MNIST. Results show that our method gets much higher IS and lower FID compared with some current GAN architectures.
Submission history
From: Sen Pei [view email][v1] Thu, 5 Aug 2021 03:41:14 UTC (43,603 KB)
[v2] Mon, 23 Aug 2021 09:28:26 UTC (37,827 KB)
[v3] Tue, 24 Aug 2021 11:52:40 UTC (39,052 KB)
[v4] Mon, 13 Sep 2021 11:46:55 UTC (34,803 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.