D2ADA: Dynamic Density-aware Active Domain Adaptation for Semantic Segmentation

TL;DR

D2ADA introduces a dynamic active domain adaptation framework for semantic segmentation that efficiently utilizes limited target domain labels by focusing on high-density target samples with low source density, improving performance with minimal annotations.

Contribution

The paper proposes a novel density-aware active learning strategy combined with a dynamic scheduling policy for domain adaptation in semantic segmentation, reducing labeling effort while maintaining high accuracy.

Findings

Outperforms existing methods on GTA5 -> Cityscapes and SYNTHIA -> Cityscapes benchmarks.

Achieves comparable results to full supervision with less than 5% target labels.

Demonstrates the effectiveness of density-aware sampling and dynamic budget adjustment.

Abstract

In the field of domain adaptation, a trade-off exists between the model performance and the number of target domain annotations. Active learning, maximizing model performance with few informative labeled data, comes in handy for such a scenario. In this work, we present D2ADA, a general active domain adaptation framework for semantic segmentation. To adapt the model to the target domain with minimum queried labels, we propose acquiring labels of the samples with high probability density in the target domain yet with low probability density in the source domain, complementary to the existing source domain labeled data. To further facilitate labeling efficiency, we design a dynamic scheduling policy to adjust the labeling budgets between domain exploration and model uncertainty over time. Extensive experiments show that our method outperforms existing active learning and domain adaptation baselines on two benchmarks, GTA5 -> Cityscapes and SYNTHIA -> Cityscapes. With less than 5% target domain annotations, our method reaches comparable results with that of full supervision. Our code is publicly available at https://github.com/tsunghan-wu/D2ADA.