UCL: Unsupervised Curriculum Learning for water body classification from remote sensing imagery

• Published in: International journal of applied earth observation and geoinformation 2021-12, Vol.105, p.102568, Article 102568
• Main Authors: Abid, Nosheen, Shahzad, Muhammad, Malik, Muhammad Imran, Schwanecke, Ulrich, Ulges, Adrian, Kovács, György, Shafait, Faisal
• Format: Article
• Language: English
• Subjects: Aircraft Imagery; Deep Learning; Machine Learning; Maskininlärning; Multi-scale Classification; Remote Sensing; Sentinel-2; Unsupervised Curriculum Learning; Water classification
• ISSN: 1569-8432; 1872-826X; 1872-826X
• DOI: 10.1016/j.jag.2021.102568

•An unsupervised deep model – UCL for water body classification from RGB remote sensing imagery.•Integration of clustering with curriculum learning leads to unsupervised learning of the deep model.•The proposed approach mitigates the data labeling requirement which requires expert knowledge.•UCL is evaluated on three different continents’ corpora; two are space-borne and one is air-bone. This paper presents a Convolutional Neural Networks (CNN) based Unsupervised Curriculum Learning approach for the recognition of water bodies to overcome the stated challenges for remote sensing based RGB imagery. The unsupervised nature of the presented algorithm eliminates the need for labelled training data. The problem is cast as a two class clustering problem (water and non-water), while clustering is done on deep features obtained by a pre-trained CNN. After initial clusters have been identified, representative samples from each cluster are chosen by the unsupervised curriculum learning algorithm for fine-tuning the feature extractor. The stated process is repeated iteratively until convergence. Three datasets have been used to evaluate the approach and show its effectiveness on varying scales: (i) SAT-6 dataset comprising high resolution aircraft images, (ii) Sentinel-2 of EuroSAT, comprising remote sensing images with low resolution, and (iii) PakSAT, a new dataset we created for this study. PakSAT is the first Pakistani Sentinel-2 dataset designed to classify water bodies of Pakistan. Extensive experiments on these datasets demonstrate the progressive learning behaviour of UCL and reported promising results of water classification on all three datasets. The obtained accuracies outperform the supervised methods in domain adaptation, demonstrating the effectiveness of the proposed algorithm.