VisCUIT: Visual Auditor for Bias in CNN Image Classifier
This addresses bias issues in CNN classifiers for practitioners and researchers, offering a novel tool to reduce manual effort in bias investigation, though it is incremental as it builds on existing visualization techniques.
The paper tackles the problem of bias in CNN image classifiers by introducing VisCUIT, an interactive visualization system that identifies underperforming subgroups and reveals image concepts causing misclassifications, resulting in a tool that runs in browsers and is open-source for easy access and extension.
CNN image classifiers are widely used, thanks to their efficiency and accuracy. However, they can suffer from biases that impede their practical applications. Most existing bias investigation techniques are either inapplicable to general image classification tasks or require significant user efforts in perusing all data subgroups to manually specify which data attributes to inspect. We present VisCUIT, an interactive visualization system that reveals how and why a CNN classifier is biased. VisCUIT visually summarizes the subgroups on which the classifier underperforms and helps users discover and characterize the cause of the underperformances by revealing image concepts responsible for activating neurons that contribute to misclassifications. VisCUIT runs in modern browsers and is open-source, allowing people to easily access and extend the tool to other model architectures and datasets. VisCUIT is available at the following public demo link: https://poloclub.github.io/VisCUIT. A video demo is available at https://youtu.be/eNDbSyM4R_4.