Discriminately Decreasing Discriminability with Learned Image Filters

TL;DR

This paper introduces an algorithm to learn image filters that enhance discriminability for a target task while reducing it for a distractor task, useful for privacy and dataset generalization.

Contribution

The paper proposes a novel method to optimize filters that selectively decrease discriminability of distractor tasks without harming the primary task.

Findings

Effective on simulated data

Successful on natural face images

Improves privacy and generalization

Abstract

In machine learning and computer vision, input images are often filtered to increase data discriminability. In some situations, however, one may wish to purposely decrease discriminability of one classification task (a "distractor" task), while simultaneously preserving information relevant to another (the task-of-interest): For example, it may be important to mask the identity of persons contained in face images before submitting them to a crowdsourcing site (e.g., Mechanical Turk) when labeling them for certain facial attributes. Another example is inter-dataset generalization: when training on a dataset with a particular covariance structure among multiple attributes, it may be useful to suppress one attribute while preserving another so that a trained classifier does not learn spurious correlations between attributes. In this paper we present an algorithm that finds optimal filters to give high discriminability to one task while simultaneously giving low discriminability to a distractor task. We present results showing the effectiveness of the proposed technique on both simulated data and natural face images.