Deep learning techniques have been widely used in autonomous driving systems for the semantic understanding of urban scenes. However, they need a huge amount of labeled data for training, which is difficult and expensive to acquire. A recently proposed workaround is to train deep networks using synthetic data, but the domain shift between real world and synthetic representations limits the performance. In this work we propose an unsupervised domain adaptation strategy to adapt a synthetic supervised training to real world data. The proposed learning strategy exploits two components: a standard supervised learning on real world data and an adversarial learning module that exploits both labeled synthetic data and unlabeled real data. Furthermore, we describe a simple method based on Fourier Transforms for unsupervised domain adaptation aimed to obtain better performances, whereby the discrepancy between the source and target distributions is reduced by swapping the low-frequency spectrum of one with the other.
for more info, I suggest you to read our report, or to check our presentation.
The main objective of this project is to become familiar with the task of Domain Adaptation applied to the Real-time Semantic Segmentation networks. The student should understand the general approaches to perform Domain Adaptation in Semantic Segmentation and the main reason to apply them to real-time networks