Class-Conditional Domain Adaptation on Semantic Segmentation
This work addresses a specific limitation in domain adaptation for semantic segmentation, making it an incremental improvement for applications requiring robust generalization across domains.
The paper tackles the problem of lower probability classes being disadvantaged in unsupervised domain adaptation for semantic segmentation by introducing a Class-Conditional Domain Adaptation method, which achieves performance comparable to state-of-the-art methods on two transfer tasks.
Semantic segmentation is an important sub-task for many applications, but pixel-level ground truth labeling is costly and there is a tendency to overfit the training data, limiting generalization. Unsupervised domain adaptation can potentially address these problems, allowing systems trained on labelled datasets from one or more source domains (including less expensive synthetic domains) to be adapted to novel target domains. The conventional approach is to automatically align the representational distributions of source and target domains. One limitation of this approach is that it tends to disadvantage lower probability classes. We address this problem by introducing a Class-Conditional Domain Adaptation method (CCDA). It includes a class-conditional multi-scale discriminator and the class-conditional loss. This novel CCDA method encourages the network to shift the domain in a class-conditional manner, and it equalizes loss over classes. We evaluate our CCDA method on two transfer tasks and demonstrate performance comparable to state-of-the-art methods.