U.S. flag

An official website of the United States government, Department of Justice.

Visualizing Deep Similarity Networks

NCJ Number
254594
Journal
2019 IEEE Winter Conference on Applications of Computer Vision (WACV) Dated: 2019 Pages: 2029-2037
Author(s)
Abby Stylianou; Richard Souvenir; Robert Pless
Date Published
2019
Length
9 pages
Annotation
For convolutional neural network models that optimize an image embedding, this article proposes a method to highlight the regions of images that contribute most to pairwise similarity.
Abstract
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. This approach is generalized to embedding networks that use different pooling strategies and provides a simple mechanism to support image similarity searches on objects or sub-regions in the query image. (publisher abstract modified)