Semi-Supervised Fine-Tuning for Deep Learning Models in Remote Sensing Applications

TL;DR

This paper introduces a semi-supervised fine-tuning approach that improves deep learning models for land cover identification in remote sensing by using SSL-enhanced loss functions, leading to better pixel-level segmentation results.

Contribution

It presents a novel combination of deep learning and semi-supervised learning for remote sensing, demonstrating improved model performance with SSL-based loss functions.

Findings

SSL-enhanced loss functions improve segmentation accuracy

Models trained with SSL approaches outperform traditional methods

Pixel-level land cover classification benefits from semi-supervised fine-tuning

Abstract

A combinatory approach of two well-known fields: deep learning and semi supervised learning is presented, to tackle the land cover identification problem. The proposed methodology demonstrates the impact on the performance of deep learning models, when SSL approaches are used as performance functions during training. Obtained results, at pixel level segmentation tasks over orthoimages, suggest that SSL enhanced loss functions can be beneficial in models' performance.