MSMatch: Semi-Supervised Multispectral Scene Classification with Few Labels
This work addresses the challenge of expensive and tedious data labeling in remote sensing by enabling competitive performance with few labels, which is incremental as it builds on semi-supervised learning but applies it to multispectral data.
The paper tackles the problem of requiring large labeled datasets for supervised learning in remote sensing by introducing MSMatch, a semi-supervised learning approach that achieves state-of-the-art results on scene classification benchmarks, with accuracies up to 19.76% better than previous methods on EuroSAT and 90.71% with just five labeled examples per class on UC Merced Land Use.
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 semi-supervised 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 achieves state-of-the-art results on EuroSAT with an accuracy that is up to 19.76% better than previous methods depending on the number of labeled training examples. With just five labeled examples per class, we reach 94.53% and 95.86% accuracy on the EuroSAT RGB and multispectral datasets, respectively. On the UC Merced Land Use dataset, we outperform previous works by up to 5.59% and reach 90.71% with five labeled examples. 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.