Biased Binary Attribute Classifiers Ignore the Majority Classes

Xinyi Zhang, Johanna Sophie Bieri, Manuel Günther

Keywords: Facial Attribute Classification, Class Activation Mapping, Binary Classifiers

TL;DR: We extend CAM techniques to binary classifiers and show that biased facial attribute classifiers do not base their majority class predictions on reasonable locations of the images.

Abstract:

To visualize the regions of interest that classifiers base their decisions on, different Class Activation Mapping (CAM) methods have been developed.
However, all of these techniques target categorical classifiers only, though most real-world tasks are binary classification.
In this paper, we extend gradient-based CAM techniques to work with binary classifiers and visualize the active regions for binary attribute classifiers.
When training an unbalanced binary classifier on a biased dataset, it is well-known that the majority class is mostly predicted much better than minority class.
In our experiments on the CelebA dataset, we verify these results, when training an unbalanced classifier to extract 40 facial attributes simultaneously.
One would expect that the biased classifier has learned to extract features mainly for the majority classes and that the proportional energy of the activations mainly reside in certain specific regions of the image where the attribute is located.
However, we find very little regular activation for samples of majority classes, while the active regions for minority classes seem mostly reasonable and overlap with our expectations.
These results suggest that biased classifiers only rely on bias activation for majority classes.
When training a balanced classifier on the unbalanced data by employing attribute-specific class weights, positive and negative classes are classified similarly well and show expected activations for almost all attributes.

Primary Area: visualization or interpretation of learned representations

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 1268