Compare and Contrast: Learning Prominent Visual Differences

TL;DR

This paper introduces a model for identifying the most noticeable differences between images, reflecting human perception, and demonstrates its effectiveness in improving image comparison tasks like search and description generation.

Contribution

It presents the concept of prominent differences, collects annotations, and develops a model that predicts these differences, enhancing image comparison methods.

Findings

Model outperforms baseline methods on UT-Zap50K and LFW10 datasets.

Prominence model improves image search and description generation tasks.

Captures human-like perception of noticeable differences.

Abstract

Relative attribute models can compare images in terms of all detected properties or attributes, exhaustively predicting which image is fancier, more natural, and so on without any regard to ordering. However, when humans compare images, certain differences will naturally stick out and come to mind first. These most noticeable differences, or prominent differences, are likely to be described first. In addition, many differences, although present, may not be mentioned at all. In this work, we introduce and model prominent differences, a rich new functionality for comparing images. We collect instance-level annotations of most noticeable differences, and build a model trained on relative attribute features that predicts prominent differences for unseen pairs. We test our model on the challenging UT-Zap50K shoes and LFW10 faces datasets, and outperform an array of baseline methods. We then demonstrate how our prominence model improves two vision tasks, image search and description generation, enabling more natural communication between people and vision systems.