Image-to-image Translation as a Unique Source of Knowledge

TL;DR

This paper investigates the effectiveness of image-to-image translation methods in transferring knowledge between dissimilar domains like SAR and optical satellite imagery, assessing how well features are transferred and how stacking improves results.

Contribution

It introduces a systematic evaluation of I2I translation in cross-domain satellite imagery and proposes stacking as a method to combine knowledge from multiple translations.

Findings

Stacking improves translation quality over single models.

Significant transfer of features from optical to SAR imagery.

Evaluation metrics show effective domain adaptation.

Abstract

Image-to-image (I2I) translation is an established way of translating data from one domain to another but the usability of the translated images in the target domain when working with such dissimilar domains as the SAR/optical satellite imagery ones and how much of the origin domain is translated to the target domain is still not clear enough. This article address this by performing translations of labelled datasets from the optical domain to the SAR domain with different I2I algorithms from the state-of-the-art, learning from transferred features in the destination domain and evaluating later how much from the original dataset was transferred. Added to this, stacking is proposed as a way of combining the knowledge learned from the different I2I translations and evaluated against single models.