Module streamline operations #736
Conversation
|
Build passed ! Good Job 🍻 ! |
|
I like the new organization and I see that all tests are still passing (and I believe all modified scripts were covered). Since this PR is just about moving around things I think the really important PR is the next one where the functions are tested, there is not much to comment on since I agree with the structure and you "only" moved code. Thank you, it look great! |
|
Thanks! I just added a small change, good to go for me! |
|
Build Failed 💥 |
EmmaRenauld
force-pushed
the
module_streamline_operations
branch
from
ef59d02 to
5f8db35
Compare
|
Build passed ! Good Job 🍻 ! |
|
Build passed ! Good Job 🍻 ! |
| @@ -123,7 +126,7 @@ def _extract_and_save_tails_heads_from_endpoints(gt_endpoints, out_dir): | |||
| affine = mask_img.affine | |||
| dimensions = mask.shape | |||
|
|
|||
| head, tail = split_heads_tails_kmeans(mask) | |||
| head, tail = split_mask_blobs_kmeans(mask, nb_clusters=2) | |||
There was a problem hiding this comment.
Why the extra parameter ? Does it change the behaviour of the function ?
There was a problem hiding this comment.
No. I just moved the function to image.utils because it only takes a mask as parameter, no tractogram. So it felt weird to keep it named split_head_tail. It could split any mask into two blobs. And why not into K blobs, if eventually anyone needs it. So it went from:
Tractanalysis.tools:
X = np.argwhere(data)
k_means = KMeans(n_clusters=2).fit(X)
mask_1 = np.zeros(data.shape)
mask_2 = np.zeros(data.shape)
mask_1[tuple(X[np.where(k_means.labels_ == 0)].T)] = 1
mask_2[tuple(X[np.where(k_means.labels_ == 1)].T)] = 1
To
image.utils:
X = np.argwhere(data)
k_means = KMeans(n_clusters=nb_clusters).fit(X)
masks = []
for i in range(nb_clusters):
mask_i = np.zeros(data.shape)
mask_i[tuple(X[np.where(k_means.labels_ == i)].T)] = 1
masks.append(mask_i)
return masks
But I forgot to change the doctstring. Doing it now, thanks.
|
Build passed ! Good Job 🍻 ! |
Next step in preparing tests for scilpy: module streamline operations.
Absolutely no line of code was changed here: only moved stuff around, clarifying the folder separations.
Now the 'streamline operation' scripts that we had identified in our previous meeting (@frheault, @AntoineTheb , @AlexVCaron) only import stuff from operation modules, nothing from tractanalysis. They will be ready for tests in a next PR.