[ECCV 2026] Delineate Anything v2: A Global Foundation Model for Field Delineation

July 25, 2026 · View on GitHub

by Mykola Lavreniuk, Nataliia Kussul, Andrii Shelestov, Yevhenii Salii, Volodymyr Kuzin, Charlotte Julia Li-Xing Wang, Zoltan Szantoi

Delineate Anything v2 extends Delineate Anything model into a globally representative, resolution-agnostic foundation model that scales agricultural field boundary detection to a planetary level from any imagery source. Trained on FBIS-73M, a massive 73-million-instance dataset spanning 61 countries with diverse imagery sources ranging from 0.25m to 10m resolution, built through a resolution-specific curation pipeline that solves the parcel-versus-field mismatch, Delineate Anything v2 sets a new state-of-the-art in global zero-shot delineation. It delivers a +103.3% relative gain in mAP@0.5 over Delineate Anything while maintaining extreme efficiency, mapping all of Ukraine (603,000 km²) in 5.4 hours on a regular PC with 1 GPU NVIDIA RTX 5070 Ti 16 GB.

intro v1

intro v2

News

2026/07/16: 🔥 Delineate Anything v2: A Global Foundation Model for Field Delineation is accepted to ECCV 2026!
2025/11/17: Delineate Anything Flow: Fast, Country-Level Field Boundary Detection from Any Source published!
2025/09/07: 🚀🚀🚀 Autobounds released for convenient field boundary detection with Delineate-Anything, directly in the browser!
👉 Demo Video | Live App.
2025/08/30: 🚀🚀 Our paper on Delineate-Anything accepted at ECAI 2025 🎉.
2025/07/07: 🚀 Delineate-Anything integrated into the TorchGeo library.

📊 Models & Performance

Global Benchmark (100-Country Independent Evaluation)

Method	mAP@0.5	mAP@0.5:0.95	Precision	Recall	Latency (ms)	Size	Download
Delineate Anything	0.275	0.103	0.345	0.454	25.0	125 MB	Download
Delineate Anything v2	0.559	0.278	0.639	0.525	25.0	125 MB	Download

Regional Performance Breakdown (mAP@0.5)

Method	Europe	Africa	Asia & Oceania	Latin America	North America
Delineate Anything	0.332	0.251	0.161	0.314	0.317
Delineate Anything v2	0.612	0.584	0.440	0.563	0.618

*Note: Both models in the global benchmark are evaluated on the newly curated manual benchmark spanning 100 countries! Original baseline evaluation from the Delineate Anything paper used FBIS-22M test splits, which primarily covered only Europe.

⚙️ Environment Setup

To set up the environment on a Linux system:

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm ~/miniconda3/miniconda.sh

source ~/miniconda3/bin/activate
conda install -c conda-forge gdal

# optional: pip install torch==2.6.0
pip install -r requirements.txt

To set up the environment on a Windows system:

conda create --prefix=./.conda python=3.11
conda activate ./.conda
conda install -c conda-forge gdal
# optional: pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128
pip install -r requirements.txt

🚀 Inference

💡 Try the Colab demo first, no installation needed, or run locally if you prefer full control.

Place your RGB images in the data/images/ folder. If available, also include the corresponding land cover map in the data/masks/ (Three Sentinel-2 sample images and a land cover map are provided for testing.)
Run the inference script:
```
python delineate.py -b batch_sample.yaml
```
The vectorized field boundaries will be saved as a GeoPackage in: data/delineated/

(Optional) To shift the resulting vector geometries:

Shift using image pixels:

python shift.py -i PATH_TO_SRC_GPKG -o PATH_TO_DST_GPKG -s PATH_TO_SAMPLE_IMAGE -x SHIFT_PIXELS_X -y SHIFT_PIXELS_Y

Shift using spatial units (SRS):

python shift.py -i PATH_TO_SRC_GPKG -o PATH_TO_DST_GPKG -x SHIFT_UNITS_X -y SHIFT_UNITS_Y

ℹ️ Tip: For advanced settings, refer to the instructions in delineation_config_guide.md

@inproceedings{lavreniuk2026delanyv2,
      title={Delineate Anything v2: A Global Foundation Model for Field Delineation}, 
      author={Mykola Lavreniuk and Nataliia Kussul and Andrii Shelestov and Yevhenii Salii and Volodymyr Kuzin and Charlotte Julia Li-Xing Wang and Zoltan Szantoi},
      year={2026},
      booktitle={European Conference on Computer Vision Workshops (ECCVW)},
}

@inproceedings{lavreniuk2025delineateanything,
      title={Delineate Anything: Resolution-Agnostic Field Boundary Delineation on Satellite Imagery}, 
      author={Mykola Lavreniuk and Nataliia Kussul and Andrii Shelestov and Bohdan Yailymov and Yevhenii Salii and Volodymyr Kuzin and Zoltan Szantoi},
      year={2025},
      booktitle={European Conference on Artificial Intelligence},
}

@article{lavreniuk2025delineateanythingflow,
      title={Delineate Anything Flow: Fast, Country-Level Field Boundary Detection from Any Source}, 
      author={Mykola Lavreniuk and Nataliia Kussul and Andrii Shelestov and Yevhenii Salii and Volodymyr Kuzin and Sergii Skakun and Zoltan Szantoi},
      year={2025},
      journal={https://arxiv.org/abs/2511.13417},
}