Unsupervised Traffic Scene Generation with Synthetic 3D Scene Graphs
This addresses the problem of generating realistic synthetic data for training autonomous driving neural networks, though it appears incremental as it builds on existing scene graph and synthesis techniques.
The paper tackles the domain gap between synthetic and real-world traffic scene images by proposing an unsupervised method that uses synthetic 3D scene graphs with spatial information to directly synthesize realistic traffic scenes without rendering, demonstrating effectiveness through scene manipulation.
Image synthesis driven by computer graphics achieved recently a remarkable realism, yet synthetic image data generated this way reveals a significant domain gap with respect to real-world data. This is especially true in autonomous driving scenarios, which represent a critical aspect for overcoming utilizing synthetic data for training neural networks. We propose a method based on domain-invariant scene representation to directly synthesize traffic scene imagery without rendering. Specifically, we rely on synthetic scene graphs as our internal representation and introduce an unsupervised neural network architecture for realistic traffic scene synthesis. We enhance synthetic scene graphs with spatial information about the scene and demonstrate the effectiveness of our approach through scene manipulation.