mariam-madata/serengeti-wildlife-classifier
Domaine:
environment and energyType de record:
model
Créateur:
mar
Hôte:
African wildlife species classification using EfficientNet transfer learning — 96.3% accuracy on buffalo, elephant, rhino, zebra
# 🦁 Serengeti African Wildlife Classifier
**Deep learning model for automated wildlife species identification — Serengeti National Park, Tanzania**
## 📊 Results
| Metric | Score |
|---|---|
| Test Accuracy | **96.3%** |
| Macro F1 Score | **0.96** |
| Training Time | ~4 minutes |
## Per-Class Performance
| Species | Precision | Recall | F1 |
|---|---|---|---|
| 🐃 Buffalo | 0.97 | 0.92 | 0.95 |
| 🐘 Elephant | 0.94 | 0.99 | 0.97 |
| 🦏 Rhino | 0.95 | 0.96 | 0.95 |
| 🦓 Zebra | 1.00 | 0.98 | 0.99 |
> Zebra achieved perfect precision (1.00)
## 🧠 Model
- EfficientNet-B0 Transfer Learning
- PyTorch framework
- 1,504 camera trap images
## 🌍 Impact
Tanzania generates $2.5B annually from wildlife tourism. This tool supports anti-poaching and conservation monitoring.
## 👤 Author
**Mariam Khamis Madata** — Dar es Salaam, Tanzania
Portfolio · GitHub · LinkedIn
## 🔗 Related Projects
Diabetic Retinopathy Detection — Medical AI · 83% accuracy