# AI4SmallFarms: Smallholder Crop Field Boundary Delineation from Sentinel-2, Vietnam and Cambodia (TACO)

## Facts

- Provider: University of Twente (ITC); TACO repackaging by ISP, Universitat de Valencia
- License: CC-BY-4.0 (https://phys-techsciences.datastations.nl/dataset.xhtml?persistentId=doi:10.17026/dans-xy6-ngg6)
- Size: 9 files
- Format: PNG (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/isp-uv-es/ai4smallfarms-taco
- Access: public, readable without an account
- DOI: https://doi.org/10.1109/LGRS.2023.3323095
- Category: Agriculture & Environment
- Source: https://huggingface.co/datasets/isp-uv-es/ai4smallfarms-taco
- Page: https://scigantic.com/archive/1a9700cf-6419-5855-aeee-d49b03a86fda
- Keywords: Agriculture & Environment, satellite imagery, benchmark, PNG, Hugging Face

## About

AI4SmallFarms (IEEE GRSL, 2023) is a benchmark for delineating agricultural field boundaries in smallholder landscapes from 10 m Sentinel-2 imagery, where fields are often smaller than a hectare. It provides 62 tiles of about 5x5 km over Vietnam and Cambodia carrying 439,001 manually digitised field polygons, which form the benchmark, and 87 tiles over the Netherlands that the authors used for pretraining. Each sample is a native-size Sentinel-2 tile with bands B2, B3, B4 and B8 (reflectance x 10,000) and a binary field-extent mask; the release's own per-region splits are kept (103 train, 22 validation, 24 test tiles). About 550 MB in TACO format, read with tacoreader. The original is deposited in the DANS Data Station for Physical and Technical Sciences under CC BY 4.0; cite the AI4SmallFarms GRSL 2023 paper.

## Load the data

```python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="isp-uv-es/ai4smallfarms-taco", filename="ai4smallfarms.zip", repo_type="dataset")
print(path)
```

## Data structure

File types in sampled files: PNG

Example file: `ai4smallfarms.zip`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/1a9700cf-6419-5855-aeee-d49b03a86fda
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/1a9700cf-6419-5855-aeee-d49b03a86fda
- Over MCP: `get_schema_card` with id `1a9700cf-6419-5855-aeee-d49b03a86fda` at https://api.scigantic.com/mcp

## Related topics

- [Satellite imagery datasets](https://scigantic.com/topics/satellite-imagery)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [PNG datasets](https://scigantic.com/topics/png)
- [Agriculture & Environment datasets](https://scigantic.com/topics/agriculture-and-environment)
