Satellite Image Time Series Classification with Pixel-Set Encoders and Temporal Self-Attention

TL;DR

This paper introduces a novel satellite image time series classification method using pixel-set encoders and temporal self-attention, outperforming previous models in accuracy and efficiency, and provides a new benchmark dataset.

Contribution

The paper presents a new neural architecture replacing convolutional layers with pixel-set encoders and using self-attention for temporal features, improving performance and resource efficiency.

Findings

Outperforms state-of-the-art in classification accuracy

Reduces processing time and memory usage

Provides a large annotated benchmark dataset

Abstract

Satellite image time series, bolstered by their growing availability, are at the forefront of an extensive effort towards automated Earth monitoring by international institutions. In particular, large-scale control of agricultural parcels is an issue of major political and economic importance. In this regard, hybrid convolutional-recurrent neural architectures have shown promising results for the automated classification of satellite image time series.We propose an alternative approach in which the convolutional layers are advantageously replaced with encoders operating on unordered sets of pixels to exploit the typically coarse resolution of publicly available satellite images. We also propose to extract temporal features using a bespoke neural architecture based on self-attention instead of recurrent networks. We demonstrate experimentally that our method not only outperforms previous state-of-the-art approaches in terms of precision, but also significantly decreases processing time and memory requirements. Lastly, we release a large open-access annotated dataset as a benchmark for future work on satellite image time series.