WHITE BLOOD CELL CLASSIFICATION USING FEDERATED LEARNING
WHITE BLOOD CELL CLASSIFICATION USING FEDERATED LEARNING
This was my course project for the subject "Linear Algebra" and I focused on classifying white blood cells (WBC) from blood smear images to support faster and more reliable medical analysis. Traditional manual methods are slow and prone to error, and while deep learning offers strong automation capabilities, patient data privacy remains a major concern. To address this, I applied a federated learning approach, enabling model training without sharing raw medical images. Both raw images and feature-extracted inputs were evaluated, and the results show that federated learning outperforms conventional deep learning. The findings also indicate that raw images provide better performance for this model, making the approach both effective and privacy-preserving.
BCCD Dataset is a small-scale dataset for blood cells detection
This dataset contains 12,500 augmented images of blood cells, there are approximately 3,000 images for each of 4 different cell types
The image is converted to HSV and the hue component is extracted
View Code: Code