MSMatch: Semisupervised Multispectral Scene Classification With Few Labels

Abstract

Supervised learning techniques are at the center of many tasks in remote sensing. Unfortunately, these methods, especially recent deep learning methods, often require large amounts of labeled data for training. Even though satellites acquire large amounts of data, labeling the data is often tedious, expensive, and requires expert knowledge. Hence, improved methods that require fewer labeled samples are needed. We present MSMatch, the first semisupervised learning approach competitive with supervised methods on scene classification on the EuroSAT and UC Merced Land Use benchmark datasets. We test both RGB and multispectral images of EuroSAT and perform various ablation studies to identify the critical parts of the model. The trained neural network outperforms previous methods by up to 19.76% and 5.59% on EuroSAT and the UC Merced Land Use datasets, respectively. With just five labeled examples per class, we attain 90.71% and 95.86% accuracy on the UC Merced Land Use dataset and EuroSAT, respectively. Our results show that MSMatch is capable of greatly reducing the requirements for labeled data. It translates well to multispectral data and should enable various applications that are currently infeasible due to a lack of labeled data. We provide the source code of MSMatch online to enable easy reproduction and quick adoption.

Keywords

• Deep learning
• multispectral image classification
• scene classification
• semisupervised learning