Delving into Rectifiers in Style-Based Image Translation
arXiv:2111.10546
Abstract
While modern image translation techniques can create photorealistic synthetic images, they have limited style controllability, thus could suffer from translation errors. In this work, we show that the activation function is one of the crucial components in controlling the direction of image synthesis. Specifically, we explicitly demonstrated that the slope parameters of the rectifier could change the data distribution and be used independently to control the direction of translation. To improve the style controllability, two simple but effective techniques are proposed, including Adaptive ReLU (AdaReLU) and structural adaptive function. The AdaReLU can dynamically adjust the slope parameters according to the target style and can be utilized to increase the controllability by combining with Adaptive Instance Normalization (AdaIN). Meanwhile, the structural adaptative function enables rectifiers to manipulate the structure of feature maps more effectively. It is composed of the proposed structural convolution (StruConv), an efficient convolutional module that can choose the area to be activated based on the mean and variance specified by AdaIN. Extensive experiments show that the proposed techniques can greatly increase the network controllability and output diversity in style-based image translation tasks.
arXiv comment14 pages,14 figures
References in corpus (11)
- MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
- A Learned Representation For Artistic Style
- Multimodal Unsupervised Image-to-Image Translation
- Semantic Image Synthesis with Spatially-Adaptive Normalization
- DRIT++: Diverse Image-to-Image Translation via Disentangled Representations
- Few-Shot Unsupervised Image-to-Image Translation
- Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression
- GAN Prior Embedded Network for Blind Face Restoration in the Wild
- Conditional Image-to-Image Translation