Hashing with binary autoencoders

Abstract

An attractive approach for fast search in image databases is binary hashing, where each highdimensional, real-valued image is mapped onto a low-dimensional, binary vector and the search is done in this binary space. Finding the optimal hash function is difficult because it involves binary constraints, and most approaches approximate the optimization by relaxing the constraints and then binarizing the result. Here, we focus on the binary autoencoder model, which seeks to reconstruct an image from the binary code produced by the hash function. We show that the optimization can be simplified with the method of auxiliary coordinates. This reformulates the optimization as alternating two easier steps: one that learns the encoder and decoder separately, and one that optimizes the code for each image. Image retrieval experiments, using precision/recall and a measure of code utilization, show the resulting hash function outperforms or is competitive with state-of-the-art methods for binary hashing.

Kernel auto-encoder for semi-supervised hashing

“…Following [4], we use the quadratic penalty method [15] for solving (4) that results in minimizing the following objective function:…”

“…Very recently, Carreira-Perpinan and Raziperchikolaei [4] have proposed a learning hash method named Binary Auto-encoder (BA) based on the well-known Autoencoder model. BA optimizes jointly over the hash functions and the binary codes using a recently proposed method of auxiliary coordinates (MAC) [5].…”

“…By applying the quadraticpenalty method [4], BA optimizes over Z and (h, f ) using the coordinate descent algorithm.…”

Unsupervised Deep Hashing With Pseudo Labels for Scalable Image Retrieval

“…Earlier unsupervised binary coding methods often use handcraft features such as GIST [1], [28] and SIFT [29], [30], which can make hashing methods achieve good performance, but still not very satisfactory for our applications. Fortunately, at near recent, on account of great development in the area of deep learning [31]- [34], hashing methods turn to extract deep features automatically using Convolutional Neural Network (CNN) descriptors [35]- [38], which inherit high discriminative property and often get state-of-the-art performance.…”

“…ITQ-CCA [22], aim to maximise the correlation between binary code and visual feature. 4) Deep Hashing avoids to use hand-crafted features such as GIST [1], SIFT [29], and make the feature extraction and encoding into a unified end-to-end model.…”