Time2Agri: Temporal Pretext Tasks for Agricultural Monitoring

TL;DR

This paper introduces three agriculture-specific self-supervised pretext tasks that leverage temporal information for improved remote sensing-based agricultural monitoring, outperforming existing methods on multiple datasets.

Contribution

It proposes three novel temporal pretext tasks tailored for agricultural landscapes, addressing a gap in existing self-supervised learning approaches for remote sensing.

Findings

Future-Frame Prediction achieves 69.6% IoU on crop mapping.

Frequency Prediction reduces yield prediction error to 30.7% MAPE.

Scaling FF to national scale yields 54.2% IoU on field boundary delineation.

Abstract

Self Supervised Learning(SSL) has emerged as a prominent paradigm for label-efficient learning, and has been widely utilized by remote sensing foundation models(RSFMs). Recent RSFMs including SatMAE, DoFA, primarily rely on masked autoencoding(MAE), contrastive learning or some combination of them. However, these pretext tasks often overlook the unique temporal characteristics of agricultural landscape, namely nature's cycle. Motivated by this gap, we propose three novel agriculture-specific pretext tasks, namely Time-Difference Prediction(TD), Temporal Frequency Prediction(FP), and Future-Frame Prediction(FF). Comprehensive evaluation on SICKLE dataset shows FF achieves 69.6% IoU on crop mapping and FP reduces yield prediction error to 30.7% MAPE, outperforming all baselines, and TD remains competitive on most tasks. Further, we also scale FF to the national scale of India, achieving 54.2% IoU outperforming all baselines on field boundary delineation on FTW India dataset.