Research map / Label-efficient and robust learning / Self-supervised and contrastive
Time2Agri: Temporal Pretext Tasks for Agricultural Monitoring
Moti Rattan Gupta, Anupam Sobti
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| pretrain with future-frame prediction instead of MAE | ViT-S · SICKLE crop type mapping | IoU: 0.67138 -> 0.69595 | Improved |
| pretrain with temporal frequency prediction instead of MAE | ViT-S · SICKLE crop yield (in-season) | MAPE: 0.32701 -> 0.30727 | Improved |
| pretrain with temporal frequency prediction instead of FF for crop type | ViT-S · SICKLE crop type mapping | IoU: 0.69595 -> 0.61976 | Hurt |
| use temporal pretext tasks vs supervised baselines on date prediction | FF, FP, TD · SICKLE sowing/transplanting/harvest date | MAPE: harvest: 0.047596 (UNet3D) vs 0.068059 (FF) | No clear effect |
| pretrain regionally (Tamil Nadu) rather than nationally (India) with FF (ablation) | FF · SICKLE crop type mapping | IoU: 0.60483 -> 0.69595 | Improved |
| pretrain regionally rather than nationally with FF (ablation) | FF · SICKLE crop yield (in-season) | MAPE: 0.37407 -> 0.32516 | Improved |
| restrict FF bitemporal pairs to exactly 3 months apart (ablation) | FF · SICKLE crop type mapping | IoU: 0.69595 -> 0.58737 | Hurt |
| use 3 and 6 month gaps for FF pairs (ablation) | FF · SICKLE crop yield (region) | MAPE: 0.36765 -> 0.33411 | Improved |
| use 3, 6 and 9 month gaps for FF pairs (ablation) | FF · SICKLE harvest date | MAPE: 0.068059 -> 0.05391 | Improved |
| use FF pretraining instead of MAE | ViT-S with supervised pretraining on France then India finetune · FTW India | IoU: 0.465626 -> 0.490628 | Improved |
| use FF pretraining instead of MAE when supervised on India only | ViT-S · FTW India (India-only) | IoU: 0.497104 -> 0.492768 | No clear effect |
| use FF instead of global RSFMs DoFA and CopernicusFM | ViT-S · SICKLE crop type mapping | IoU: 0.58363 / 0.57229 -> 0.69595 | Improved |
Closest papers by meaning
- Can Large Multimodal Models Understand Agricultural Scenes? Benchmarking with AgroMind (NeurIPS 2025 · similarity 0.86)
- Exact: Exploring Space-Time Perceptive Clues for Weakly Supervised Satellite Image Time Series Semantic Segmentation (CVPR 2025 · similarity 0.86)
- Fields of The World: A Machine Learning Benchmark Dataset for Global Agricultural Field Boundary Segmentation (AAAI 2025 · similarity 0.85)
- Combining Deep Learning and Street View Imagery to Map Smallholder Crop Types (AAAI 2024 · similarity 0.85)
- AgroOmni: A Large-Scale Multi-view Agricultural Dataset for Cross-Scale Multimodal Reasoning (NeurIPS 2026 · similarity 0.83)
- HarvestNet: A Dataset for Detecting Smallholder Farming Activity Using Harvest Piles and Remote Sensing (AAAI 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
