Hello,
in the test submission wiki document it says: "Please ensure that all prediction files have the same image shape size as the original image. (The output is an integer array with the same shape as the input image.)"
For the cine validation submission, the results had to be submitted as
a) Transposed arrays, i.e. the order of the axes switched compared to the input. So, if
Input:
```
h5py.File('MultiCoil/Cine/TrainingSet/AccFactor04/P001/cine_lax.mat')['kspace_sub04'].shape
Out: (12, 3, 10, 204, 448) #(time, view, coil, undersampled, fullysampled dimension)
```
we had to supply the results as
Output:
```
:h5py.File('test/MultiCoil/Cine/ValidationSet/AccFactor04/P001/cine_sax.mat' ) ['img4ranking']
Out: (171, 123, 2, 3) # (fullysampled dimension, undersampled dimension, view, time)
```
b) The resulting array was supposed to be floats. Not integer.
So, will the test results our docker container has to provide be judged like to validation results (floats, transposed order of the axis, just not cropped)
Or will a different script be used and have to supply the order of the axes differently?
Would it maybe be possible to just provide
```
ds=h5py.File(WORKING_EXAMPLE_FILE.mat)['img4ranking']
ds.shape, ds.dtype, np.max(ds)
```
to all teams?
Created by Felix Zimmermann fzimmermann89 The evaluation code (testing phase) will not be released before the announcement of the final results.
But your can refer to the evaluation code during validation phase, which is almost the same. Please do all the procedures the same as validation phase except for cropping the images. Also, can you tell us where we can find the script that will be used for evaluation? Is it somewhere on github?
Thank you :)