TY  - CHAP
U1  - Konferenzveröffentlichung
A1  - Durall Lopez, Ricard
A1  - Pfreundt, Franz-Josef
A1  - Keuper, Janis
ED  - Farinella, Giovanni Maria
ED  - Radeva, Petia
ED  - Braz, Jose
ED  - Bouatouch, Kadi
T1  - Stabilizing GANs with Octave Convolutions
T2  - Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications
N2  - In this preliminary report, we present a simple but very effective technique to stabilize the training of CNN based GANs. Motivated by recently published methods using frequency decomposition of convolutions (eg Octave Convolutions), we propose a novel convolution scheme to stabilize the training and reduce the likelihood of a mode collapse. The basic idea of our approach is to split convolutional filters into additive high and low frequency parts, while shifting weight updates from low to high during the training. Intuitively, this method forces GANs to learn low frequency coarse image structures before descending into fine (high frequency) details. Our approach is orthogonal and complementary to existing stabilization methods and can simply plugged into any CNN based GAN architecture. First experiments on the CelebA dataset show the effectiveness of the proposed method.
KW  - Generative Adversarial Network
KW  - Octave Convolution
KW  - Stability
KW  - Regularization
Y1  - 2021
UR  - https://www.scitepress.org/Papers/2021/101787/101787.pdf
SN  - 2184-4321
SS  - 2184-4321
SN  - 978-989-758-488-6
SB  - 978-989-758-488-6
U6  - https://doi.org/10.5220/0010178700150023
DO  - https://doi.org/10.5220/0010178700150023
VL  - 4
SP  - 15
EP  - 23
PB  - SciTePress
ER  -