Time2Agri: Temporal Pretext Tasks for Agricultural Monitoring

2025-07-06

AI Summary

Existing agricultural remote sensing self-supervised methods neglect the natural temporal cycles of crop growth, hindering effective modeling of spatiotemporal dynamics. To address this, we propose—tailored to agro-landscape characteristics—the first set of customized time-aware self-supervised pretraining tasks: temporal difference prediction, time-frequency prediction, and future-frame prediction. Our approach integrates masked autoencoding with explicit temporal modeling to systematically capture crop phenological cycles. Evaluated on the SICKLE dataset, our method achieves 69.6% IoU in future-frame prediction (crop mapping) and 30.7% MAPE in time-frequency prediction (yield estimation). At national scale, validation on FTW India yields 54.2% IoU. This work establishes a novel time-aware self-supervised paradigm explicitly designed for agricultural phenology modeling, significantly advancing downstream task performance in crop mapping and yield forecasting.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

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.

Problem

• Addresses lack of temporal focus in agricultural SSL tasks
• Proposes novel pretext tasks for crop and yield prediction
• Improves performance on field boundary delineation nationally

Innovation

• Time-Difference Prediction for agricultural cycles
• Temporal Frequency Prediction for yield accuracy
• Future-Frame Prediction for crop mapping