Realistic Image Synthesis with Configurable 3D Scene Layouts
This addresses the challenge for users in image synthesis who need rich controllability over 3D geometric aspects, representing an incremental improvement over existing conditional image synthesis approaches.
The paper tackles the problem of generating realistic images with precise control over 3D geometry, such as object positions and orientations, by proposing a method that uses configurable 3D scene layouts to synthesize images with geometrically correct structures and supports manipulations like viewpoint changes and style adjustments.
Recent conditional image synthesis approaches provide high-quality synthesized images. However, it is still challenging to accurately adjust image contents such as the positions and orientations of objects, and synthesized images often have geometrically invalid contents. To provide users with rich controllability on synthesized images in the aspect of 3D geometry, we propose a novel approach to realistic-looking image synthesis based on a configurable 3D scene layout. Our approach takes a 3D scene with semantic class labels as input and trains a 3D scene painting network that synthesizes color values for the input 3D scene. With the trained painting network, realistic-looking images for the input 3D scene can be rendered and manipulated. To train the painting network without 3D color supervision, we exploit an off-the-shelf 2D semantic image synthesis method. In experiments, we show that our approach produces images with geometrically correct structures and supports geometric manipulation such as the change of the viewpoint and object poses as well as manipulation of the painting style.