---
language: de
license: apache-2.0
tags:
- image-classification
- computer-vision
- birds
- ornithology
- wildlife
- efficientnet
datasets:
- custom
metrics:
- accuracy
model-index:
- name: german-bird-classifier
  results:
  - task:
      type: image-classification
      name: Image Classification
    metrics:
    - type: accuracy
      value: 100.0
      name: Test Accuracy
---

> ⚠️ **DEPRECATED - Use v2 for Better Accuracy**
> 
> This is **version 1** of the German Bird Classifier with 87.69% accuracy.
> 
> **Please use [german-bird-classifier-v2](https://huggingface.co/kamera-linux/german-bird-classifier-v2) instead:**
> - ✅ **99.71% accuracy** (vs. 87.69% in v1)
> - ✅ **42× fewer errors**
> - ✅ **Perfect classification** for 5 out of 8 species
> 
> This v1 model will remain available for compatibility but is no longer actively maintained.
> 
> ---

- accuracy
model-index:
- name: german-bird-classifier
  results:
  - task:
      type: image-classification
      name: Image Classification
    metrics:
    - type: accuracy
      value: 100.0
      name: Test Accuracy
---

# German Bird Classifier 🐦

A fine-tuned EfficientNet-B0 model for classifying 8 common German garden bird species with **100% validation accuracy**.

## Model Description

This model was trained using [vogel-model-trainer](https://github.com/kamera-linux/vogel-model-trainer), a toolkit for creating bird classification models from video footage.

**Base Model:** `google/efficientnet-b0` (8.5M parameters)  
**Fine-tuned for:** German garden birds classification  
**Training Framework:** PyTorch + Hugging Face Transformers

## Supported Species

The model can classify the following 8 bird species:

1. **Blaumeise** (Blue Tit) - *Cyanistes caeruleus*
2. **Grünling** (European Greenfinch) - *Chloris chloris*
3. **Haussperling** (House Sparrow) - *Passer domesticus*
4. **Kernbeißer** (Hawfinch) - *Coccothraustes coccothraustes*
5. **Kleiber** (Eurasian Nuthatch) - *Sitta europaea*
6. **Kohlmeise** (Great Tit) - *Parus major*
7. **Rotkehlchen** (European Robin) - *Erithacus rubecula*
8. **Sumpfmeise** (Marsh Tit) - *Poecile palustris*

## Performance

| Species | Validation Accuracy | Samples |
|---------|---------------------|---------|
| Blaumeise | 100.0% | 5 |
| Grünling | 100.0% | 5 |
| Haussperling | 100.0% | 5 |
| Kernbeißer | 100.0% | 5 |
| Kleiber | 100.0% | 5 |
| Kohlmeise | 100.0% | 5 |
| Rotkehlchen | 100.0% | 5 |
| Sumpfmeise | 100.0% | 5 |
| **Overall** | **100.0%** | **40** |

## Usage

### With vogel-model-trainer

```bash
# Install the toolkit
pip install vogel-model-trainer

# Extract birds from video using this classifier
vogel-trainer extract --folder ~/bird-data \
  --species-model kamera-linux/german-bird-classifier \
  --sample-rate 20 --skip-blurry --deduplicate \
  video.mp4
```

### With Python

```python
from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import torch

# Load model and processor
model = AutoModelForImageClassification.from_pretrained("kamera-linux/german-bird-classifier")
processor = AutoImageProcessor.from_pretrained("kamera-linux/german-bird-classifier")

# Load and preprocess image
image = Image.open("bird.jpg")
inputs = processor(images=image, return_tensors="pt")

# Predict
with torch.no_grad():
    outputs = model(**inputs)
    logits = outputs.logits
    predicted_class = logits.argmax(-1).item()

# Get species name
species = model.config.id2label[predicted_class]
print(f"Predicted species: {species}")
```

## Training Details

- **Training Date:** November 13, 2025
- **Optimizer:** AdamW with cosine LR schedule
- **Augmentation:** Random rotation, affine, color jitter, gaussian blur
- **Regularization:** Weight decay 0.01, label smoothing 0.1
- **Early Stopping:** 7 epochs patience
- **Input Size:** 224x224 pixels
- **Batch Size:** 32
- **Learning Rate:** 2e-5

## Dataset

The model was trained on custom-collected video footage of German garden birds using the vogel-model-trainer toolkit:
- Video sources: Garden bird feeders in Germany
- Extraction method: YOLO-based bird detection + quality filtering
- Data split: 80% training / 20% validation
- Preprocessing: Blur detection, deduplication, class balancing

## Limitations

- Trained specifically on German garden birds - may not generalize to other species or regions
- Best performance on clear, well-lit images similar to training data
- May struggle with:
  - Juvenile birds with different plumage
  - Birds in flight or unusual poses
  - Heavy occlusion or poor lighting
  - Species not in the 8 training classes

## Ethical Considerations

This model is intended for:
- Wildlife monitoring and conservation
- Educational purposes
- Citizen science projects
- Automated bird feeder cameras

**Not recommended for:**
- Commercial wildlife tracking without proper permits
- Any use that could harm bird populations

## Citation

If you use this model, please cite:

```bibtex
@software{german_bird_classifier_2025,
  author = {Kamera Linux},
  title = {German Bird Classifier},
  year = {2025},
  url = {https://huggingface.co/kamera-linux/german-bird-classifier},
  note = {Trained with vogel-model-trainer: https://github.com/kamera-linux/vogel-model-trainer}
}
```

## Links

- **Training Toolkit:** [vogel-model-trainer on GitHub](https://github.com/kamera-linux/vogel-model-trainer)
- **PyPI Package:** [vogel-model-trainer](https://pypi.org/project/vogel-model-trainer/)
- **Model Repository:** [kamera-linux/german-bird-classifier](https://huggingface.co/kamera-linux/german-bird-classifier)

## License

This model is released under the **Apache License 2.0**. See the [LICENSE](https://huggingface.co/kamera-linux/german-bird-classifier/blob/main/LICENSE) file for details.

## Acknowledgments

Built with:
- [Hugging Face Transformers](https://huggingface.co/transformers/)
- [PyTorch](https://pytorch.org/)
- [Google EfficientNet](https://github.com/google/automl)
- [Ultralytics YOLO](https://github.com/ultralytics/ultralytics)