search

assassint2017 / MICCAI-LITS2017

liver segmentation using deep learning
388 stars 91 forks source link

what will be the input image? #4

Open ahmadmubashir opened 5 years ago

ahmadmubashir commented 5 years ago

should we give the input the whole 3D CT images? please explain it the folder name 'train/ct' and 'train/seg' please also explain it.

Thanks

ahmadmubashir commented 5 years ago

do we need any per processing? I prepare the dataset with your code. get_random_data.py then i make the 3d patches of size 323232 and given it to the network. but loss is not decreasing. please help me. thanks

mitiandi commented 5 years ago

should we give the input the whole 3D CT images? please explain it the folder name 'train/ct' and 'train/seg' please also explain it.

Thanks

should we give the input the whole 3D CT images? please explain it the folder name 'train/ct' and 'train/seg' please also explain it.

Thanks

Hi, the author uses 3D CT images as the input of network,but they should be pre-processed. 'train/ct' means the ct volumes of train set, while 'train/seg' means their ground truths. Best wishes to you!

ahmadmubashir commented 5 years ago

should we give the input the whole 3D CT images? please explain it the folder name 'train/ct' and 'train/seg' please also explain it. Thanks

should we give the input the whole 3D CT images? please explain it the folder name 'train/ct' and 'train/seg' please also explain it. Thanks

Hi, the author uses 3D CT images as the input of network,but they should be pre-processed. 'train/ct' means the ct volumes of train set, while 'train/seg' means their ground truths. Best wishes to you!

Thank you. But for training, we need to give the whole volume? Or distribute it to 3D multiple equal size patches? This confuses me.

mitiandi commented 5 years ago

https://github.com/assassint2017/MICCAI-LITS2017/issues/5#issuecomment-434156126