About Test

[Dasha484]

When I run test.py, an error is reported in the console. The traceback is as follows:
Traceback (most recent call last):
File "test.py", line 121, in
main()
File "test.py", line 94, in main
original_affine = batch["label_meta_dict"]["affine"][0].numpy()
KeyError: 'label_meta_dict'
How should I fix this?Additionally, how can I obtain the segmentation accuracy for each organ in the BTCV dataset? Looking forward to your response.
This is the "val" part of my dataset.json.
"validation": [
{
"image": "imagesTr/img0035.nii.gz",
"label": "labelsTr/label0035.nii.gz"
},
{
"image": "imagesTr/img0036.nii.gz",
"label": "labelsTr/label0036.nii.gz"
},
{
"image": "imagesTr/img0037.nii.gz",
"label": "labelsTr/label0037.nii.gz"
},
{
"image": "imagesTr/img0038.nii.gz",
"label": "labelsTr/label0038.nii.gz"
},
{
"image": "imagesTr/img0039.nii.gz",
"label": "labelsTr/label0039.nii.gz"
},
{
"image": "imagesTr/img0040.nii.gz",
"label": "labelsTr/label0040.nii.gz"
}
]
Looking forward to your response.

[zhangxuan-thecastle]

文件“test.py”，第 121 行，在
main（） 文件“test.py”，第 94 行，在主
original_affine = batch[“label_meta_dict”][“affine”][0].numpy（）

KeyError：“label_meta_dict”

I have also encountered this problem. How can I solve “label_meta_dict” and “affine”

[ericspod]

The issue with the key error may be related to changes to LoadImaged such that it doesn't produce the metadata dictionary any more. Wherever LoadImaged occurs add the argument image_only=False to produce the dictionary. For the metric, there would have to be quite significant changes to how the dice is calculated here, the MONAI DiceMetric class would allow per-class values to be calculated. @ahatamiz or @upupming I think you worked on this as well? Thanks

[Bigsealion]

The issue with the key error may be related to changes to LoadImaged such that it doesn't produce the metadata dictionary any more. Wherever LoadImaged occurs add the argument image_only=False to produce the dictionary. For the metric, there would have to be quite significant changes to how the dice is calculated here, the MONAI DiceMetric class would allow per-class values to be calculated. @ahatamiz or @upupming I think you worked on this as well? Thanks

Hello, may I ask which version of MONAI corresponds to the current swinUNETR? I'm unable to install it from the requirements.txt. Thank you!

[ASCI02]

The issue with the key error may be related to changes to LoadImaged such that it doesn't produce the metadata dictionary any more. Wherever LoadImaged occurs add the argument image_only=False to produce the dictionary. For the metric, there would have to be quite significant changes to how the dice is calculated here, the MONAI DiceMetric class would allow per-class values to be calculated. @ahatamiz or @upupming I think you worked on this as well? Thanks

Hello, may I ask which version of MONAI corresponds to the current swinUNETR? I'm unable to install it from the requirements.txt. Thank you!
You can install monai version 1.1.0 with 'pip install 'monai[all]==1.1.0'', which works for swinunetr