search

Shrey-Viradiya / Pneumonia_Detection

This repository contains code for pneumonia detection using X-ray images of the lungs.
https://shrey-viradiya.github.io/blog/Pneumonia_Detection/
MIT License
10 stars 3 forks source link

Dataset Labels #5

Closed Shrey-Viradiya closed 4 years ago

Shrey-Viradiya commented 4 years ago

Our dataset has nested labels. If it is normal then okay, but if it is cases of pneumonia then whether it is bacteria or virus. If virus then COVID or other.

look at the table: Label Label_1_Virus_category Label_2_Virus_category Image_Count
Normal - - 1576
Pnemonia Stress-Smoking ARDS 2
Pnemonia Virus - 1493
Pnemonia Virus COVID-19 58
Pnemonia Virus SARS 4
Pnemonia bacteria 2772
Pnemonia bacteria Streptococcus 5

data.zip

Shrey-Viradiya commented 4 years ago

Data File is included...data.zip

tirth-hihoriya commented 4 years ago

We have only 58 datapoints of Pnemonia with Virus labeled COVID-19. And few category like ARDS, SARS, Streptococccus have only 2, 4 and 5 datapoints. Do you have any idea to tackle it? @Shrey-Viradiya @tirthasheshpatel @Tirth1306

Shrey-Viradiya commented 4 years ago

Shall I use only COVID-19 images and let other images go or add some extra output neurons to predict all the classes? @tirthasheshpatel

tirthasheshpatel commented 4 years ago

add some extra output neurons to predict all the classes

This seems better but I haven't looked at the images. If images of all classes look alike, it will be better to predict all the classes and avoid overfitting. If the images are significantly different, we may purge other class off in training.

Shrey-Viradiya commented 4 years ago

I've skipped Streptococcus, ARDS, and SARS. I kept only Normal, Viral, Bacterial and Covid-19