Creating Ensembles of Classifiers through UMDA for Aerial Scene Classification

TL;DR

This work tackles the challenge of aerial scene classification by systematically evaluating six Deep Metric Learning losses across four pre-trained CNN backbones. It demonstrates that deep metric learning can outperform traditional pre-trained CNN baselines on AID, UCMerced, and RESISC45, with ensemble fusion via UMDA further enhancing performance by selectively combining roughly half of the available classifiers. The study provides a comprehensive diversity analysis showing complementary information among DML+DLA pairs and confirms that UMDA-based ensembles yield consistent gains across datasets, achieving relative improvements over the best single tuple and PT-CNN baselines. Overall, the results highlight the value of combining deep metric learning with evolutionary ensemble methods for robust and scalable remote sensing scene classification.

Abstract

Aerial scene classification, which aims to semantically label remote sensing images in a set of predefined classes (e.g., agricultural, beach, and harbor), is a very challenging task in remote sensing due to high intra-class variability and the different scales and orientations of the objects present in the dataset images. In remote sensing area, the use of CNN architectures as an alternative solution is also a reality for scene classification tasks. Generally, these CNNs are used to perform the traditional image classification task. However, another less used way to classify remote sensing image might be the one that uses deep metric learning (DML) approaches. In this sense, this work proposes to employ six DML approaches for aerial scene classification tasks, analysing their behave with four different pre-trained CNNs as well as combining them through the use of evolutionary computation algorithm (UMDA). In performed experiments, it is possible to observe than DML approaches can achieve the best classification results when compared to traditional pre-trained CNNs for three well-known remote sensing aerial scene datasets. In addition, the UMDA algorithm proved to be a promising strategy to combine DML approaches when there is diversity among them, managing to improve at least 5.6% of accuracy in the classification results using almost 50\% of the available classifiers for the construction of the final ensemble of classifiers.

Figures (4)

• Figure 1: Difference between traditional classification and metric learning approaches.
• Figure 2: Examples of aerial scenes from UCMerced dataset used in this work.
• Figure 3: Correlation Coefficient ($\rho$) as a diversity measurement computed among $24$ classifiers (DML$+$DLA) in a 5-folds cross validation protocol using AID dataset. A similar behavior occurred as using the UCMerced and RESISC45 datasets.
• Figure 4: UMDA and MV accuracy results (%) for each of the five folds of the AID dataset.