Mostefa-Terbeche/diabetic-retinopathy-deepdrid-resnet50

Diabetic Retinopathy Classification Model

Model Description

This model is trained for diabetic retinopathy classification using the resnet50 architecture on the deepdrid dataset.

Model Details

• Architecture: resnet50
• Dataset: deepdrid
• Training Date: 20250618-000522
• Task: 5-class diabetic retinopathy grading (0-4)

Performance

• Test Accuracy: 0.8375
• Test Quadratic Kappa: 0.9584199584199584
• Validation Kappa: 0.9584199584199584

Usage

import torch
from huggingface_hub import hf_hub_download

# Download model
model_path = hf_hub_download(
    repo_id="your-username/diabetic-retinopathy-deepdrid-resnet50",
    filename="model_best.pt"
)

# Load model
model = torch.load(model_path, map_location='cpu')

Classes

• 0: No DR (No diabetic retinopathy)
• 1: Mild DR (Mild non-proliferative diabetic retinopathy)
• 2: Moderate DR (Moderate non-proliferative diabetic retinopathy)
• 3: Severe DR (Severe non-proliferative diabetic retinopathy)
• 4: Proliferative DR (Proliferative diabetic retinopathy)

Citation

If you use this model, please cite your research paper/thesis.