Paddy Land Segmentation and Estimation Using U-Net and SAM with Satellite Imagery: A Sri Lankan Study
Bibliographic record
Abstract
Monitoring crop cultivation is crucial for ensuring food security, effective land-use planning, and sustainable agriculture. Accurate mapping and measurement of cultivated and non-cultivated paddy lands are critical for developing effective agricultural plans and climate-resilient farming in Sri Lanka. This study aims to develop automated, deep learning-based segmentation models using high-resolution satellite images from Google Earth Pro to segment and estimate the area of paddy fields. A comprehensive methodology was employed, involving the collection of satellite images, preprocessing, manual annotation of segmentation masks, model training, and performance evaluation. We employed four segmentation strategies: (1) distinguishing cultivated fields from the background, (2) separating non-cultivated fields from the background, (3) combining cultivated and non-cultivated fields against the background, and (4) a multi-class model segmenting background, cultivated, and non-cultivated areas simultaneously. Fine-tuned U-Net and the Segment Anything Model (SAM) were applied separately to these four strategies. SAM achieved a Dice coefficient above 90% in the first three cases, while U-Net outperformed SAM in the fourth strategy with a Dice coefficient of 91%. The pixel-based analysis was done in the Manthai East division, Northern Province, Sri Lanka, to determine the actual area estimation and verify the real-world relevance of this model. The proposed segmentation models demonstrate that deep learning has the potential to segment crops accurately and offers a cost-effective approach for monitoring agricultural land use practices. This research also paves the way for estimating paddy lands for agrarian field officers using freely available high-resolution satellite images for their purposes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".