| --- |
| license: cc-by-nc-4.0 |
| task_categories: |
| - image-segmentation |
| tags: |
| - agriculture |
| - remote-sensing |
| - uav |
| - crop |
| - breeding |
| - polygon |
| pretty_name: Plot2Phenome Preview |
| size_categories: |
| - n<1K |
| --- |
| |
| # Plot2Phenome Preview Dataset |
|
|
| **This is a preview subset** of the Plot2Phenome dataset, intended for |
| pre-publication release. The full dataset will be made available upon paper |
| acceptance. |
|
|
| ## Overview |
|
|
| Plot2Phenome is a multi-modal UAV remote sensing dataset for breeding plot |
| segmentation. Each sample consists of an RGB image with polygon annotations |
| delineating individual field plots. This preview contains **150 patches** |
| randomly sampled (5 per scene × date) from the full dataset. |
|
|
| ### Scenes |
|
|
| | Scene | Description | |
| |-------|-------------| |
| | 01_2020_rugao_wheat | Rugao, Jiangsu — Winter wheat (2020) | |
| | 02_2020_suining_wheat | Suining, Jiangsu — Winter wheat (2020) | |
| | 03_2024_rugao_rice | Rugao, Jiangsu — Rice (2024) | |
| | 04_2025_yandu_wheat | Yandu, Jiangsu — Winter wheat (2025) | |
| | 05_2025_dengzhou_maize | Dengzhou, Henan — Summer maize (2025) | |
| | 06_2025_rugao_rice | Rugao, Jiangsu — Rice (2025) | |
|
|
| ## Dataset Structure |
|
|
| ``` |
| plot2phenome_preview/ |
| data.yaml # YOLO-seg dataset config |
| images/ |
| train/ # RGB patches (512×512 PNG) |
| val/ # RGB patches (512×512 PNG) |
| labels/ |
| train/ # YOLO-seg polygon labels (normalized [0,1]) |
| val/ # YOLO-seg polygon labels (normalized [0,1]) |
| ``` |
|
|
| ## Label Format |
|
|
| Ultralytics YOLO-seg format. Each ``.txt`` file corresponds to one image. |
| Each line encodes one instance polygon: |
|
|
| ``` |
| class_id x1 y1 x2 y2 ... xn yn |
| ``` |
|
|
| - ``class_id`` is always ``0`` (plot). |
| - Coordinates are normalized to ``[0, 1]`` relative to image width and height. |
| - The polygon is closed (first and last points may be identical). |
|
|
| ## Splits |
|
|
| Patches are split by flight date (train/val) to avoid spatial leakage. Each |
| scene contains patches from multiple growth stages captured on different dates. |
|
|
| ## Usage with Ultralytics |
|
|
| ```python |
| from ultralytics import YOLO |
| |
| model = YOLO('yolov8x-seg.pt') |
| model.train(data='data.yaml', epochs=100, imgsz=512) |
| ``` |
|
|
| ## License |
|
|
| This preview dataset is released under |
| [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). |
|
|
| ## Citation |
|
|
| If you use this dataset, please cite our paper (forthcoming). |
|
|
| ## Full Dataset |
|
|
| The complete Plot2Phenome dataset includes multi-modal data (RGB + DSM height), |
| additional scenes, and higher sample density. It will be released at |
| [here](https://huggingface.co/datasets/little-king/Plot2Phenome) upon paper acceptance. |
|
|
