Classification Drives Geographic Bias in Street Scene Segmentation

TL;DR

This paper investigates how geographic biases in training data affect street scene segmentation models, revealing that classification errors contribute significantly to geo-biases and can be mitigated by coarser class grouping.

Contribution

It demonstrates that classification errors are a primary source of geo-bias in segmentation models and proposes a simple mitigation strategy using coarser classes.

Findings

Eurocentric models are geo-biased in segmentation tasks.

Classification errors account for 10-90% of geo-biases.

Using coarser classes reduces geo-biases significantly.

Abstract

Previous studies showed that image datasets lacking geographic diversity can lead to biased performance in models trained on them. While earlier work studied general-purpose image datasets (e.g., ImageNet) and simple tasks like image recognition, we investigated geo-biases in real-world driving datasets on a more complex task: instance segmentation. We examined if instance segmentation models trained on European driving scenes (Eurocentric models) are geo-biased. Consistent with previous work, we found that Eurocentric models were geo-biased. Interestingly, we found that geo-biases came from classification errors rather than localization errors, with classification errors alone contributing 10-90% of the geo-biases in segmentation and 19-88% of the geo-biases in detection. This showed that while classification is geo-biased, localization (including detection and segmentation) is geographically robust. Our findings show that in region-specific models (e.g., Eurocentric models), geo-biases from classification errors can be significantly mitigated by using coarser classes (e.g., grouping car, bus, and truck as 4-wheeler).