Skip to content

andreacate/ANNDL_Homework1

Repository files navigation

ANNDL_Homework1

Blood Cell Classification Project

Overview

This project leverages deep learning techniques to classify blood cell images into eight distinct types, including basophils, neutrophils, lymphocytes, and more. The final model achieves 93% accuracy on the test set, using EfficientNetV2S as a feature extractor for image classification. This project demonstrates the application of convolutional neural networks (CNNs) for medical image classification, showcasing the power of deep learning in the healthcare domain.

Dataset

  • 13,759 RGB images (96x96 pixels each)
  • 8 cell type classifications, including basophils, neutrophils, lymphocytes, and others

Project Structure

The project is organized into multiple Jupyter notebooks, each serving a specific purpose:

  • Model Implementation Notebooks: Separate notebooks for each tested deep learning architecture (e.g., EfficientNetV2S, CNN).
  • Dataset Manipulation Notebook: A notebook detailing the preprocessing steps applied to the dataset, including data augmentation, normalization, and encoding of labels.

About

Homework on multi-class classification problem

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published