Generative Adversarial Networks using Adaptive Convolution
This work addresses a specific bottleneck in GAN architectures for image generation, offering incremental improvements in modeling visual diversity.
The paper tackled the problem of fixed upsampling operations in GANs limiting their ability to model diverse visual appearances by proposing an adaptive convolution method that learns upsampling based on local context. Experiments on CIFAR-10 and STL-10 datasets showed large improvements over baselines and achieved state-of-the-art performance in unsupervised settings.
Most existing GANs architectures that generate images use transposed convolution or resize-convolution as their upsampling algorithm from lower to higher resolution feature maps in the generator. We argue that this kind of fixed operation is problematic for GANs to model objects that have very different visual appearances. We propose a novel adaptive convolution method that learns the upsampling algorithm based on the local context at each location to address this problem. We modify a baseline GANs architecture by replacing normal convolutions with adaptive convolutions in the generator. Experiments on CIFAR-10 dataset show that our modified models improve the baseline model by a large margin. Furthermore, our models achieve state-of-the-art performance on CIFAR-10 and STL-10 datasets in the unsupervised setting.