Week 2: Blood Cell Classification
Team Members: Emre Tunç, Muhammed Sezer Berker, Tolga Furkan Güler
Theme: Classifying the blood cell images and predict the possible disease according to the blood cell we have detected.
Reminder
Our aim in this project is correctly classify the blood cell images of the 4 classes and predict the disease that the person possessing this cell.
We will use Blood Cell Images dataset. This dataset contains 12,500 augmented images of blood cells.There are approximately 3,000 images for each of 4 different cell types grouped into 4 different folders.This week we worked on the preprocessing of the dataset to be used in the cnn model.
Data Preprocessing
Our dataset is divided into 3 part as test, validation and train.When we examined the distribution of our data set, we obtained a graph like this.
We found that the distribution of the dataset was appropriate for the training process.No uneven distribution between classes.
When we examined our dataset, we found that Image data augmentation technique was applied to the images.This technique is obtained different versions of images in the dataset by applying different augmentation techniques.In this way, we can artificially enlarge our data, it helps the model learn better by making variations in the technical images and prevents the underfitting.
We have converted all of the photographs to gray scale format. The main reason for doing this is to overcome the problems caused by the multidimensional photographs during the training phase. When the photographs are used in multidimensional mode, the number of required parameters will increase and the training will take longer.
This week, we have made our dataset suitable for training. Thank you for reading and for your time.See you next week.
