# Visualizing Deep Similarity Networks

## TL;DR

This paper introduces a visualization method for deep similarity networks that highlights image regions influencing pairwise similarity, aiding understanding and supporting similarity searches on objects or sub-regions.

## Contribution

It extends visualization tools to similarity networks, showing how they focus on different features and enabling image similarity searches based on regions or objects.

## Key findings

- Visualization reveals feature focus differences in similarity networks.
- Method supports similarity searches on objects or sub-regions.
- Applicable to various pooling strategies in embedding networks.

## Abstract

For convolutional neural network models that optimize an image embedding, we propose a method to highlight the regions of images that contribute most to pairwise similarity. This work is a corollary to the visualization tools developed for classification networks, but applicable to the problem domains better suited to similarity learning. The visualization shows how similarity networks that are fine-tuned learn to focus on different features. We also generalize our approach to embedding networks that use different pooling strategies and provide a simple mechanism to support image similarity searches on objects or sub-regions in the query image.