Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes
This work addresses the challenge of training convolutional neural networks for semantic segmentation in autonomous driving applications by reducing reliance on costly real-world annotations through synthetic data adaptation, though it is incremental in nature.
The paper tackles the domain mismatch problem in semantic segmentation for urban scenes by proposing a curriculum-style learning approach that first learns global and local label distributions from the target domain, then regularizes predictions to follow these properties, resulting in outperformance over baselines on two datasets and two backbone networks.
During the last half decade, convolutional neural networks (CNNs) have triumphed over semantic segmentation, which is one of the core tasks in many applications such as autonomous driving. However, to train CNNs requires a considerable amount of data, which is difficult to collect and laborious to annotate. Recent advances in computer graphics make it possible to train CNNs on photo-realistic synthetic imagery with computer-generated annotations. Despite this, the domain mismatch between the real images and the synthetic data cripples the models' performance. Hence, we propose a curriculum-style learning approach to minimize the domain gap in urban scenery semantic segmentation. The curriculum domain adaptation solves easy tasks first to infer necessary properties about the target domain; in particular, the first task is to learn global label distributions over images and local distributions over landmark superpixels. These are easy to estimate because images of urban scenes have strong idiosyncrasies (e.g., the size and spatial relations of buildings, streets, cars, etc.). We then train a segmentation network while regularizing its predictions in the target domain to follow those inferred properties. In experiments, our method outperforms the baselines on two datasets and two backbone networks. We also report extensive ablation studies about our approach.