Error on data transform when training large dirty data

[ChristianEschen]

Hi.

I am training a model on a large dataset of dicom files containing "dirty data".

Somewhere in the training phase I receive an error about wrong number of channels.

Since I have 300 000 images I cannot make any manual quality assesment of the images.

is there any effective way to create a debug report of images (with paths or more informaton) causing errors?

best regards
Christian

[wyli]

thanks for reporting this, there are some discussions about extending the DataAnalyzer module to go through the datasets and perhaps detect this type of input image issue #5168

(currently it can only report image level statistics:

MONAI/monai/apps/auto3dseg/data_analyzer.py

Lines 88 to 101 in 70c0443

	from monai.apps.auto3dseg.data_analyzer import DataAnalyzer
	datalist = {
	"testing": [{"image": "image_003.nii.gz"}],
	"training": [
	{"fold": 0, "image": "image_001.nii.gz", "label": "label_001.nii.gz"},
	{"fold": 0, "image": "image_002.nii.gz", "label": "label_002.nii.gz"},
	{"fold": 1, "image": "image_001.nii.gz", "label": "label_001.nii.gz"},
	{"fold": 1, "image": "image_004.nii.gz", "label": "label_004.nii.gz"},
	],
	}
	dataroot = '/datasets' # the directory where you have the image files (nii.gz)
	DataAnalyzer(datalist, dataroot)

)