cremi / cremi_python Goto Github PK

Confidences is not imported and I cannot find it in the repository. Not sure if this is relevant, though.
Best,
Fabian

Thank you for sharing your code.
And I found that the codes of vi in skimage and cremi are also based on gala's code.
*gala:https://github.com/janelia-flyem/gala/blob/master/gala/evaluate.py
*skimage: https://github.com/scikit-image/scikit-image/blob/master/skimage/metrics/_variation_of_information.py
*cremi: https://github.com/cremi/cremi_python/blob/master/cremi/evaluation/voi.py
But there is a tiny question: Where and why to ignore 0 in VI?
For your code, gala, you ignore 0 both in gt and pred (Default)
For cremi code, they only ignore 0 in gt, not in pred (Deafult)
For skimage, they did not ignore 0 in gt and pred (Default)
I made a simple comparison below, and i found that different setting will draw out different result. Can you tell what should i do in the task of neuron segmentation? Do i need to ignore 0 label in both gt and pred?

Hi. Sorry for the dummy question. But still, I am quite confused. Take a specific region as an example, which in reconstruction the label is a, while in ground-truth is b and a != b. Can I use code to evaluate directly? Or if not, what should I do so that I can use it? Thanks.

The 51 slice in sample_A+_20160601.hdf and 88 slice in sample_A+_padded_20160601.hdf are total zeros

At these lines, two terms in the two sums are very different scales. The first is O(number of pixels squared), the second is in the range 0-1. Perhaps there are missing parentheses?

https://github.com/cremi/cremi_python/blob/master/cremi/evaluation/rand.py#L60-L61

(cc @iarganda - similar code appears in http://brainiac2.mit.edu/SNEMI3D/sites/default/files/SNEMI3D_metrics.m)

I'm having trouble squaring the comments in the code and the code itself.

In this line, it says that background (zero) pixels are excluded from the computation:
https://github.com/cremi/cremi_python/blob/master/cremi/evaluation/rand.py#L13

In this line, the total number of pixels (zero and nonzero) are computed as "n":
https://github.com/cremi/cremi_python/blob/master/cremi/evaluation/rand.py#L42

And "n" is used here in the computation:
https://github.com/cremi/cremi_python/blob/master/cremi/evaluation/rand.py#L60-L61

I would expect n to be something like "np.count_nonzero(segA)"