Mikael Brudfors
Mikael Brudfors
Hello, Could you please tell me how to best convert one of your RGB encoded label images of shape (nx, ny, 3) to a categorical image of shape (nx, ny),...
@balbasty would it be possible to add a simple demo, demonstrating fitting a U-net using nitorch?
Bugs: * Currently, jittering the resampling grid introduces a small misalignment in the resulting registration (therefore disabled by default). Is something wrong with the implementation perhaps? Enhancements: * Work on...
Sorry! Note: there will be a native matrix exponential in PyTorch 1.7, which is in pre-release state I think. Once it's released, we could dispatch expm to either our or...
GPU memory will become an issue for a large number of observed images, support reading writing from disk, directly into GPU memeory, would alleviate this problem.
For example, by segmenting background a computing its std.
This could allow for integration into, for example, some type of U-net segmentation network. Remember to test with gradcheck.
The LocalNet in the [3D registration tutorial](https://github.com/Project-MONAI/tutorials/blob/main/3d_registration/paired_lung_ct.ipynb) is currently instantiated as: ```py model = LocalNet( spatial_dims=3, in_channels=2, out_channels=3, num_channel_initial=32, extract_levels=[3], out_activation=None, out_kernel_initializer="zeros", ).to(device) ``` However, the number of extraction levels...
This PR allows for use of the metrics in `pairwise_measures.py` to be used from MONAI, in particular it enables the use of a batch dimension and a smoothing parameter. Some...
* To decrease code repetition and chance of implementation errors some functions/classes can be moved to a utility library: https://github.com/csudre/MetricsReloaded/blob/mikael-mods/utils.py and imported when needed, i.e., https://github.com/csudre/MetricsReloaded/blob/f7f6296468b6556c882cccebb8b6a973d2352e01/pairwise_measures.py#L6 This could also helps...