Bias in Automated Image Colorization: Metrics and Error Types
This work addresses bias in automated image colorization for computer vision applications, but it is incremental as it focuses on measurement and categorization without proposing new methods.
The paper measured color shifts in images colorized by the DeOldify model on the ADE20K dataset, identifying effects like desaturation, a blue shift, and category-specific variations, with manual categorization of errors into three classes.
We measure the color shifts present in colorized images from the ADE20K dataset, when colorized by the automatic GAN-based DeOldify model. We introduce fine-grained local and regional bias measurements between the original and the colorized images, and observe many colorization effects. We confirm a general desaturation effect, and also provide novel observations: a shift towards the training average, a pervasive blue shift, different color shifts among image categories, and a manual categorization of colorization errors in three classes.