search

mundher / local-global

Official implementation of Lung Nodule Classification using Deep Local-Global Networks using PyTorch
MIT License
23 stars 17 forks source link

lidc_img #3

Closed cpuxzb closed 3 years ago

cpuxzb commented 4 years ago

1.The dataset is from lidc-irdi,the lidc_img.zip has ten folders,what's the diferences of the ten folders? 2.Data Augmentation:you obtained 15 different images altogether of each nodule through the data augmentation procedure.But,there are 17 images for every 848 nodules？Why？

mundher commented 4 years ago

1.The dataset is from lidc-irdi,the lidc_img.zip has ten folders,what's the diferences of the ten folders?

All folders contain the same images. However, in every fold the training and the validation set are changing and stored in labels.csv.

2.Data Augmentation:you obtained 15 different images altogether of each nodule through the data augmentation procedure.But,there are 17 images for every 848 nodules？Why？

for every nodule, we have 18 images. check Augmenter class in preprocessing.py to understand how these images are generated