Image Retrieval System based on a Binary Auto-Encoder and a Convolutional Neural Network

The amount of image content on the Internet has increased dramatically in recent years; its precise search and retrieval is a challenge at present. The methods that have shown high efficiency are those based on convolution neural networks (CNN) and, particularly, binary coding methods based on hashing functions. This article presents a new image retrieval scheme based on attributes from a CNN, an efficient low-dimensional binary auto-encoder, and, finally, a near-neighbor retrieval stage. The proposed methodology was tested with two image datasets CIFAR-10 and MNIST. The results are compared with existing methods in the literature.