Machine Learning-Based Normal White Blood Classification and Segmentation

Project Code :TCMAPY2096

Objective

This project develops a machine learning-based system for white blood cell segmentation and multi-class classification.It uses two datasets: one with segmentation masks and another with labeled images for four cell types: Eosinophil, Lymphocyte, Monocyte, and Neutrophil.Segmentation is performed using Attention U-Net, SegFormer, and DeepLabV3 to extract accurate cell regions.Classification employs ResNeSt, ConvNeXt, and MobileNetV3-Large for efficient and precise cell type prediction.Preprocessing, augmentation, and model training ensure high performance and generalization.A Flask-based interface provides a simple workflow for image upload, segmentation, classification, and result display.The system integrates modern deep learning techniques into a unified, user-friendly platform, reducing manual effort and improving accuracy.

Abstract

This project focuses on the development of a machine learning system designed to perform white blood cell segmentation and multi-class classification. Two datasets are used: one containing cell images with masks for segmentation tasks, and another providing labeled images for classifying four categories—Eosinophil, Lymphocyte, Monocyte, and Neutrophil. The segmentation stage employs Attention U-Net, SegFormer, and DeepLabV3 to extract precise structural regions. The classification stage utilizes ResNeSt, ConvNeXt, and MobileNetV3-Large to identify cell types efficiently. A Flask-based interface is created with modules for user registration, login, segmentation, classification, and logout. The aim is to design a simplified, optimized, and accurate system that supports cell-type differentiation through deep learning workflows. The project structure ensures clarity in model usage, dataset handling, and interface design, offering a unified approach for segmentation and classification tasks using modern machine learning techniques.

Keywords: segmentation, Attention U-Net, SegFormer, DeepLabV3, classification, ResNeSt, ConvNeXt, MobileNetV3-Large, Flask, white blood cells

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.