Resampling Augmentation for Time Series Contrastive Learning: Application to Remote Sensing
This addresses the problem of leveraging abundant unlabeled remote sensing time series data for researchers in agricultural monitoring, though it appears incremental as it builds on existing contrastive learning frameworks with a new augmentation technique.
The paper tackles the challenge of designing effective data augmentations for contrastive learning in time series, specifically for Satellite Image Time Series (SITS), by introducing a novel resampling-based augmentation strategy that generates positive pairs through upsampling and extracting disjoint subsequences. The method outperforms common alternatives on agricultural classification benchmarks and achieves state-of-the-art performance on the S2-Agri100 dataset without using spatial information or temporal encodings.
Given the abundance of unlabeled Satellite Image Time Series (SITS) and the scarcity of labeled data, contrastive self-supervised pretraining emerges as a natural tool to leverage this vast quantity of unlabeled data. However, designing effective data augmentations for contrastive learning remains challenging for time series. We introduce a novel resampling-based augmentation strategy that generates positive pairs by upsampling time series and extracting disjoint subsequences while preserving temporal coverage. We validate our approach on multiple agricultural classification benchmarks using Sentinel-2 imagery, showing that it outperforms common alternatives such as jittering, resizing, and masking. Further, we achieve state-of-the-art performance on the S2-Agri100 dataset without employing spatial information or temporal encodings, surpassing more complex masked-based SSL frameworks. Our method offers a simple, yet effective, contrastive learning augmentation for remote sensing time series.