Learning a Complete Image Indexing Pipeline
This work addresses the need for scalable image retrieval systems by integrating learning into both indexing components, potentially improving efficiency and accuracy.
The paper tackles the problem of building a complete image indexing pipeline by proposing a neural framework that learns both the inverted file index and the approximate distance computation, which were previously handled separately with unsupervised and supervised methods.
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.