Add dataset card

README.md

@@ -0,0 +1,63 @@
+---
+license: mit
+language:
+- en
+tags:
+- agriculture
+- crop-disease
+- wheat
+- rice
+- computer-vision
+- plant-pathology
+pretty_name: AgriDrone Crop Disease Dataset
+size_categories:
+- 10K<n<100K
+---
+
+# AgriDrone — Crop Disease Detection Dataset
+
+Full dataset collection used for training the AgriDrone crop-disease detection
+system (21-class YOLOv8n-cls classifier, 15 wheat + 6 rice diseases).
+
+## Subfolders
+
+| Folder | Contents | Size |
+|---|---|---|
+| `raw/wheat/` | Raw wheat disease images (multi-source) | ~6.3 GB |
+| `raw/rice/` | Raw rice disease images | ~1.0 GB |
+| `raw/roboflow/` | Roboflow-sourced detection datasets | ~1.5 GB |
+| `training/` | Train/val/test splits used for the published model | ~5.2 GB |
+| `wheat_raw/` + `wheat_annotated/` | Pseudo-annotated wheat imagery | ~2.8 GB |
+| `externals/PDT_datasets/` | External benchmark: Plant Disease Treatment | ~6.1 GB |
+| `externals/pdt_plant_disease/` | External plant-disease benchmark | ~1.9 GB |
+| `working/wheat-raw/` | Current working copy (raw) | ~960 MB |
+| `working/wheat-clean/` | Current working copy (cleaned) | ~40 MB |
+| `working/wheat-split/` | Current working copy (train/val/test) | ~45 MB |
+
+## Usage
+
+```python
+from huggingface_hub import snapshot_download
+# Pull the training splits only
+path = snapshot_download(
+    repo_id="ashu010/agridrone-data",
+    repo_type="dataset",
+    allow_patterns="training/**",
+)
+```
+
+Or clone the full repo:
+
+```bash
+git lfs install
+git clone https://huggingface.co/datasets/ashu010/agridrone-data
+```
+
+## Citation
+
+See the AgriDrone repo: https://github.com/Ashut0sh-mishra/agri-drone
+
+## License
+
+MIT for the curation layer. Individual source datasets retain their original
+licenses — check each subfolder for upstream attribution.