DGInStyle: Domain-Generalizable Semantic Segmentation with Image Diffusion Models and Stylized Semantic Control
This work addresses the challenge of data scarcity and domain shift in autonomous driving perception, offering an incremental but effective solution for improving semantic segmentation models.
The authors tackled the problem of generating diverse, domain-generalizable data for semantic segmentation in autonomous driving by using a pretrained latent diffusion model with stylized semantic control, resulting in consistent performance improvements across multiple datasets compared to previous state-of-the-art methods.
Large, pretrained latent diffusion models (LDMs) have demonstrated an extraordinary ability to generate creative content, specialize to user data through few-shot fine-tuning, and condition their output on other modalities, such as semantic maps. However, are they usable as large-scale data generators, e.g., to improve tasks in the perception stack, like semantic segmentation? We investigate this question in the context of autonomous driving, and answer it with a resounding "yes". We propose an efficient data generation pipeline termed DGInStyle. First, we examine the problem of specializing a pretrained LDM to semantically-controlled generation within a narrow domain. Second, we propose a Style Swap technique to endow the rich generative prior with the learned semantic control. Third, we design a Multi-resolution Latent Fusion technique to overcome the bias of LDMs towards dominant objects. Using DGInStyle, we generate a diverse dataset of street scenes, train a domain-agnostic semantic segmentation model on it, and evaluate the model on multiple popular autonomous driving datasets. Our approach consistently increases the performance of several domain generalization methods compared to the previous state-of-the-art methods. The source code and the generated dataset are available at https://dginstyle.github.io.