SEAN: Image Synthesis with Semantic Region-Adaptive Normalization
This work addresses the need for fine-grained style control in image generation for applications like interactive editing, though it is incremental over existing normalization-based approaches.
The authors tackled the problem of controlling individual semantic region styles in image synthesis conditioned on segmentation masks, achieving better reconstruction quality, variability, and visual quality than previous methods, with improved quantitative metrics like FID and PSNR.
We propose semantic region-adaptive normalization (SEAN), a simple but effective building block for Generative Adversarial Networks conditioned on segmentation masks that describe the semantic regions in the desired output image. Using SEAN normalization, we can build a network architecture that can control the style of each semantic region individually, e.g., we can specify one style reference image per region. SEAN is better suited to encode, transfer, and synthesize style than the best previous method in terms of reconstruction quality, variability, and visual quality. We evaluate SEAN on multiple datasets and report better quantitative metrics (e.g. FID, PSNR) than the current state of the art. SEAN also pushes the frontier of interactive image editing. We can interactively edit images by changing segmentation masks or the style for any given region. We can also interpolate styles from two reference images per region.