Сегментация и типирование лейкоцитов методами глубокого обучения

Abstract

In this work, we presents preliminary results of data processing pipeline that performs segmentation and classification of the white blood cells. The data processing pipeline includes a segmentation step and a classification step. The first step is performed by a neural network with UNet architecture, and the second step is performed by the convolutional neural network. Quality of classification reaches 0.55 on the validation dataset in the categorical accuracy metric.