Learning to Generate Training Datasets for Robust Semantic Segmentation
This addresses robustness issues in safety-critical applications like autonomous driving, where reliable perception is crucial, but it is incremental as it builds on existing generative and segmentation methods.
The paper tackles the problem of improving semantic segmentation robustness to real-world perturbations and unseen object types by proposing Robusta, a robust conditional GAN that generates perturbed training images, resulting in significantly enhanced robustness against distribution shifts and out-of-distribution samples.
Semantic segmentation methods have advanced significantly. Still, their robustness to real-world perturbations and object types not seen during training remains a challenge, particularly in safety-critical applications. We propose a novel approach to improve the robustness of semantic segmentation techniques by leveraging the synergy between label-to-image generators and image-to-label segmentation models. Specifically, we design Robusta, a novel robust conditional generative adversarial network to generate realistic and plausible perturbed images that can be used to train reliable segmentation models. We conduct in-depth studies of the proposed generative model, assess the performance and robustness of the downstream segmentation network, and demonstrate that our approach can significantly enhance the robustness in the face of real-world perturbations, distribution shifts, and out-of-distribution samples. Our results suggest that this approach could be valuable in safety-critical applications, where the reliability of perception modules such as semantic segmentation is of utmost importance and comes with a limited computational budget in inference. We release our code at https://github.com/ENSTA-U2IS-AI/robusta.