This study investigates the potential of unsupervised domain adaptation (UDA) methods in improving the accuracy and fairness of skin lesion classification. The researchers leverage multiple skin lesion datasets to assess the effectiveness of UDA in binary and multi-class skin lesion classification. They find that UDA is effective in binary classification, especially when imbalance is mitigated. However, its performance is less prominent in multi-class tasks, indicating the need to address imbalance to achieve above-baseline accuracy. The study also reveals that UDA can effectively reduce bias against minority groups and promote fairness, even without the explicit use of fairness-focused techniques.
Publication date: July 6, 2023
Project Page: N/A
Paper: https://arxiv.org/pdf/2307.03157.pdf
