Classification of Bloodcells Subtypes

Basics

This is a Deep Learning based project where the task is to classify the subtype of Bloodcells from the given image using deep learning techniques.

Project Description

This dataset contains 12,500 augmented images of blood cells (JPEG) with accompanying cell type labels. The cell types are Eosinophil, Lymphocyte, Monocyte, and Neutrophil. We use different pre-trained models named Resnet50, Modified Resnet50, EfficientNetb0, Xception to find the most accurate model. Furthermore, for explainibility we use Grad-CAM.

Used

Datasets used: https://www.kaggle.com/datasets/paultimothymooney/blood-cells

Models:

• Resnet50 - Test Accuracy 88.21%
• Modified Resnet50 - Test Accuracy 89.42%
• EfficientNetb0 - Test Accuracy 77.48%
• Xception - Test Accuracy 84.64%