← Back to Projects Diabetic Retinopathy Screening with Transfer Learning
Classify retinal fundus images with CNNs and transfer learning to support diabetic retinopathy screening workflows.
Categories
CVDeep Learning
Tech Used
CNNAlexNetGoogLeNetResNetFine-tuningData processingGitfirebase authenticationFastAPIHTML/CSS
Problem
Retinal image review is time-intensive, and real-world images vary in quality, contrast, and disease visibility.
Approach
- Applied preprocessing, contrast enhancement, and augmentation to improve input robustness
- Built CNN and transfer-learning models with fine-tuning across established architectures
- Used class-balancing and threshold-aware evaluation to support screening-oriented classification
Results
- Well-structured medical imaging classification pipeline
- Stronger robustness for low-quality or visually challenging retinal samples
Demo Videos