FREDOM: Fairness Domain Adaptation Approach to Semantic Scene Understanding

Although Domain Adaptation in Semantic Scene Segmentation has shown impressive improvement in recent years, the fairness concerns in the domain adaptation have yet to be well defined and addressed. In addition, fairness is one of the most critical aspects when deploying the segmentation models into human-related real-world applications, e.g., autonomous driving, as any unfair predictions could influence human safety. In this paper, we propose a novel Fairness Domain Adaptation (FREDOM) approach to semantic scene segmentation. In particular, from the proposed formulated fairness objective, a new adaptation framework will be introduced based on the fair treatment of class distributions. Moreover, to generally model the context of structural dependency, a new conditional structural constraint is introduced to impose the consistency of predicted segmentation. Thanks to the proposed Conditional Structure Network, the self-attention mechanism has sufficiently modeled the structural information of segmentation. Through the ablation studies, the proposed method has shown the performance improvement of the segmentation models and promoted fairness in the model predictions. The experimental results on the two standard benchmarks, i.e., SYNTHIA $\to$ Cityscapes and GTA5 $\to$ Cityscapes, have shown that our method achieved State-of-the-Art (SOTA) performance.

PDF Abstract CVPR 2023 PDF CVPR 2023 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Domain Adaptation GTA5 to Cityscapes FREDOM - Transformer mIoU 73.6 # 6
Domain Adaptation GTA5 to Cityscapes FREDOM - DeepLabV2 mIoU 61.3 # 14
Domain Adaptation SYNTHIA-to-Cityscapes FREDOM - Transformer mIoU 67 # 5
Domain Adaptation SYNTHIA-to-Cityscapes FREDOM - DeepLabV2 mIoU 59.1 # 12

Methods


No methods listed for this paper. Add relevant methods here