Learning a Complete Image Indexing Pipeline

TL;DR

This paper introduces a unified neural framework for image indexing that learns both the inverted file index and approximate distance computation, improving scalability and efficiency in large-scale image retrieval.

Contribution

It presents the first system to jointly learn both components of image indexing using structured binary encoding within a neural framework.

Findings

Achieves efficient large-scale image retrieval

Unifies index learning and distance computation

Demonstrates improved retrieval accuracy

Abstract

To work at scale, a complete image indexing system comprises two components: An inverted file index to restrict the actual search to only a subset that should contain most of the items relevant to the query; An approximate distance computation mechanism to rapidly scan these lists. While supervised deep learning has recently enabled improvements to the latter, the former continues to be based on unsupervised clustering in the literature. In this work, we propose a first system that learns both components within a unifying neural framework of structured binary encoding.