A data-centric approach to class-specific bias in image data augmentation
This work addresses bias mitigation in computer vision for practitioners, but it is incremental as it builds on existing data augmentation research.
The study tackled class-specific bias in image data augmentation by analyzing its effects across various datasets and models, finding that Vision Transformers are more robust than residual models, and refined a method to reduce computational demands by a factor of 16.2.
Data augmentation (DA) enhances model generalization in computer vision but may introduce biases, impacting class accuracy unevenly. Our study extends this inquiry, examining DA's class-specific bias across various datasets, including those distinct from ImageNet, through random cropping. We evaluated this phenomenon with ResNet50, EfficientNetV2S, and SWIN ViT, discovering that while residual models showed similar bias effects, Vision Transformers exhibited greater robustness or altered dynamics. This suggests a nuanced approach to model selection, emphasizing bias mitigation. We also refined a "data augmentation robustness scouting" method to manage DA-induced biases more efficiently, reducing computational demands significantly (training 112 models instead of 1860; a reduction of factor 16.2) while still capturing essential bias trends.