search

mobie / platybrowser-project

7 stars 5 forks source link

Cell 16053 version 1.0.1 #47

Closed vzinche closed 4 years ago

vzinche commented 4 years ago

Screenshot from 2020-03-10 09-50-10

vzinche commented 4 years ago

The same with cells 19799, 21099, 10419, 10462 Basically, most of the cells that I've proofread

constantinpape commented 4 years ago

@vzinche that should be fixed now, can you please check for some of your cells (note that the cell ids have probably changed compared the previous version 1.0.1)

vzinche commented 4 years ago

The 3d connectivity seems to be fixed, but I found a bug in the symmetric cell remapping :) I guess it happens when cell 1 has a label A, cell 2 has label 0, and they both get remapped to the same cell C in the new version, the cell C gets label 0, not A (or maybe it just gets the label of the last cell?). Anyways, an example is a cell 27641 in 1.0.1 remapped from 27628 and 29253. 27628 has 'pair_index' 124, 29253 - 0, so 27641 got 0 as well, while it should have gotten 124. Additionally, we should either change all the None's in pair_index column to 0 or vice versa, because they mean the same.

vzinche commented 4 years ago

I hope this was clear :D

constantinpape commented 4 years ago

I hope this was clear :D

Partially :D.

From what I understand, the issue is that for cells that were separate in some version, then got (falsely) merged and then got split again, the mapping is incorrect. Is that right?

vzinche commented 4 years ago

Ahaha, sorry, no. They were falsely split, and correctly merged in the new version. But the resulting cell got the label 0. The problem is : when two cells get merged, how is the new label calculated?

constantinpape commented 4 years ago

Ahaha, sorry, no. They were falsely split, and correctly merged in the new version. But the resulting cell got the label 0. The problem is : when two cells get merged, how is the new label calculated?

Ah ok, I see. Let's quickly discuss after the coffee meeting.

constantinpape commented 4 years ago

Ok, I looked into this and indeed, the mapping for objects that result from a merge was more or less random (I think it just got assigned whatever the higher original id had). This should be fixed now and I checked that it works for the id you posted here. Could you check for some other ids of symmetric pairs, @vzinche?

vzinche commented 4 years ago

Yup, seems to have worked, thanks!