How poorvikanp/ai-driven-diabetic-retinopathy-detection works
This is a notebook-based diabetic-retinopathy image-classification project. Its documented workflow loads retinal images, preprocesses and splits them, trains CNN and EfficientNetB3 models, evaluates performance, and visualizes predictions with Grad-CAM. The notebook is the only implementation file shown; the README documents the stages and capabilities, but the notebook contents were not provided, so internal calls and data flow cannot be verified. The README also mentions saved models, though no model directory appears in the supplied file tree.
Main entry points: DR_Classification_Notebook (2).ipynb EfficientNetB3 transfer learning
Components 12
Image data
- Fundus images
- Dataset loading
DR_Classification_Notebook (2).ipynb - Preprocessing and augmentation
DR_Classification_Notebook (2).ipynb - Train-validation split
DR_Classification_Notebook (2).ipynb
Model training
- Simple CNN
DR_Classification_Notebook (2).ipynb - EfficientNetB3 transfer learning
DR_Classification_Notebook (2).ipynb - Saved models
Evaluation and interpretation
- Model evaluation
DR_Classification_Notebook (2).ipynb - Model comparison
DR_Classification_Notebook (2).ipynb - Prediction visualization
DR_Classification_Notebook (2).ipynb - Grad-CAM explanations
DR_Classification_Notebook (2).ipynb
Other components
- Project user
Connections 13
- Project user → Dataset loading: runs notebook
README.md - Fundus images → Dataset loading: provides images
README.md - Dataset loading → Preprocessing and augmentation: precedes
README.md - Preprocessing and augmentation → Train-validation split: precedes
README.md - Train-validation split → Simple CNN: trains
README.md - Train-validation split → EfficientNetB3 transfer learning: trains
README.md - Simple CNN → Model evaluation: evaluated by
README.md - EfficientNetB3 transfer learning → Model evaluation: evaluated by
README.md - Model evaluation → Model comparison: compared in
README.md - EfficientNetB3 transfer learning → Prediction visualization: supports
README.md - EfficientNetB3 transfer learning → Grad-CAM explanations: explained by
README.md - Simple CNN → Saved models: saved as
README.md - EfficientNetB3 transfer learning → Saved models: saved as
README.md