# Enhancing Remote Sensing Image Retrieval with Triplet Deep Metric Learning Network

## TL;DR

This paper introduces a triplet deep metric learning CNN for remote sensing image retrieval, which improves accuracy by learning a semantic feature space where similar images are closer, outperforming existing methods.

## Contribution

The paper proposes a novel triplet deep metric learning approach for remote sensing image retrieval, including supervised and unsupervised dimensionality reduction techniques, with comprehensive experimental validation.

## Key findings

- Significantly outperforms state-of-the-art retrieval methods.
- Effective semantic feature space for image similarity comparison.
- Both supervised and unsupervised dimensionality reduction improve performance.

## Abstract

With the rapid growing of remotely sensed imagery data, there is a high demand for effective and efficient image retrieval tools to manage and exploit such data. In this letter, we present a novel content-based remote sensing image retrieval method based on Triplet deep metric learning convolutional neural network (CNN). By constructing a Triplet network with metric learning objective function, we extract the representative features of the images in a semantic space in which images from the same class are close to each other while those from different classes are far apart. In such a semantic space, simple metric measures such as Euclidean distance can be used directly to compare the similarity of images and effectively retrieve images of the same class. We also investigate a supervised and an unsupervised learning methods for reducing the dimensionality of the learned semantic features. We present comprehensive experimental results on two publicly available remote sensing image retrieval datasets and show that our method significantly outperforms state-of-the-art.