---
license: mit
metrics:
- accuracy
base_model:
- google/efficientnet-b0
pipeline_tag: image-classification
tags:
- medical
- biology
---
# Diabetic Retinopathy Model

**Model Information:**
- **Architecture:** EfficientNet-B0
- **Task:** Multi-class classification (5 severity levels)
- **Dataset:** [Diabetic Retinopathy Dataset](https://www.kaggle.com/datasets/sovitrath/diabetic-retinopathy-224x224-2019-data)
- **Input Size:** 224×224 RGB images

**Classes:**
1. No_DR (No Diabetic Retinopathy)
2. Mild
3. Moderate
4. Severe
5. Proliferate_DR

**Performance Metrics:**
- **Accuracy:** 98.55%
- **Precision:** 0.9861
- **Recall:** 0.9855
- **F1-Score:** 0.9856

**Usage:**
```python
from shifaa.vision import VisionModelFactory

model = VisionModelFactory.create_model(
    model_type="classification",
    model_name="Diabetic_Retinopathy"
)

result = model.run("fundus_image.jpg", show_image=True)
print(f"Severity: {result['predicted_class']}")
print(f"Confidence: {result['confidence']:.2f}%")
```

**Confusion Matrix:**

![Confusion Matrix](./DR_CM.png)


**Preprocessing:**
- Resize to 224×224
- Random horizontal flip (training)
- Random rotation ±10° (training)
- Normalize: mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]

**Training Details:**
- **Loss Function:** CrossEntropyLoss
- **Optimizer:** Adam (lr=0.001)
- **Batch Size:** 64
- **Epochs:** 30
- **Device:** CUDA/CPU

---