# Unsupervised Neural Quantization for Compressed-Domain Similarity Search

## TL;DR

This paper introduces an unsupervised deep neural network architecture utilizing multi-codebook quantization for efficient visual descriptor compression and similarity search, significantly outperforming existing methods in large-scale image retrieval.

## Contribution

The paper presents a novel deep learning-based unsupervised quantization method that improves compression and retrieval efficiency over shallow architectures.

## Key findings

- Outperforms previous state-of-the-art quantization methods on multiple datasets.
- Enables fast data encoding and efficient distance computation.
- Demonstrates significant accuracy improvements in large-scale image retrieval.

## Abstract

We tackle the problem of unsupervised visual descriptors compression, which is a key ingredient of large-scale image retrieval systems. While the deep learning machinery has benefited literally all computer vision pipelines, the existing state-of-the-art compression methods employ shallow architectures, and we aim to close this gap by our paper. In more detail, we introduce a DNN architecture for the unsupervised compressed-domain retrieval, based on multi-codebook quantization. The proposed architecture is designed to incorporate both fast data encoding and efficient distances computation via lookup tables. We demonstrate the exceptional advantage of our scheme over existing quantization approaches on several datasets of visual descriptors via outperforming the previous state-of-the-art by a large margin.