---
license: mit
library_name: deepforest
pipeline_tag: image-classification
tags:
  - deepforest
  - crop-model
  - tree-genus
  - ecology
  - neon
---

# Tree Genus Classification (CropModel)

Classifies tree crowns detected by [DeepForest](https://github.com/weecology/DeepForest) into 54 genera. Trained on RGB imagery from 29 [NEON](https://www.neonscience.org/) sites across North America.

Trained with [NeonTreeClassification](https://github.com/GatorSense/NeonTreeClassification).

## Usage

```python
from deepforest import main
from deepforest.model import CropModel

detector = main.deepforest()
detector.load_model("weecology/deepforest-tree")

genus_model = CropModel.load_model("weecology/cropmodel-tree-genus")

results = detector.predict_tile(path="tile.tif", crop_model=genus_model)
# results has columns: cropmodel_label, cropmodel_score
```

## Results (Test Set)

| Metric | Value |
|---|---|
| Accuracy | 44.0% |
| Macro F1 | 0.25 |
| Weighted F1 | 0.44 |
| Classes | 54 |

Full per-class precision/recall/F1 in [`classification_report.csv`](classification_report.csv).

## Training

| Parameter | Value |
|---|---|
| Architecture | ResNet-18 (torchvision, ImageNet pretrained) |
| Input | 224x224 RGB, ImageNet normalization |
| Resize interpolation | nearest-neighbor |
| Optimizer | AdamW (lr=2.5e-4, weight_decay=1e-4) |
| Scheduler | ReduceLROnPlateau |
| Max epochs | 500 (early stopping patience=15) |
| Best epoch | 3 (val_loss=2.22) |
| Batch size | 256 |
| Class weights | sqrt inverse-frequency |
| Seed | 42 |

## Dataset

16,348 deduplicated tree crowns from 29 NEON sites. One sample per unique individual, rare species (<6 samples) removed. Labels from NEON Vegetation Structure Taxonomy (VST) field surveys. RGB crown crops extracted at 0.1m resolution.

| Split | Samples |
|---|---|
| Train (70%) | 11,443 |
| Val (15%) | 2,452 |
| Test (15%) | 2,453 |

Split method: stratified random, seed=42.

**Sites**: ABBY, BART, BONA, CLBJ, DEJU, DELA, GRSM, GUAN, HARV, HEAL, JERC, KONZ, LENO, MLBS, MOAB, NIWO, ONAQ, OSBS, PUUM, RMNP, SCBI, SERC, SJER, SOAP, TALL, TEAK, UKFS, UNDE, WREF

## License

MIT