SaibotMagd

Update: I tested the idea i posted before and the result seems a lot better, but now it doesn't finally close the regions (i.e. some gaps on top and at...

> Could you let me know what commands you ran and any specific setting? I haven't actually seen the type of image from your first post either and wonder how...

Update: I tried another 2D signed symmetrical image:  same result as before. - unsigned 2D and 3D doesn't work - signed 3D doesn't work - signed 2D does work...

> The underlying cause is most likely that the edge algorithm used here doesn't handle 2D images or thin/irregular 3D structures so well. This is something to improve. This seems...

So after I used a bspline registration on the source plates to create a more smooth dataset it creates nice edges for the annotation as you saw here: https://github.com/sanderslab/magellanmapper/issues/45 But...

> Thanks for posting this. So the refinement worked well when using horizontal sections, but not coronal planes? Yes as I showed in #45 . I did not try coronal...

The edge refinement in coronal direction seems similar as in horizontal, so this isn't a real problem right now:  The result of the refinement step shows the main problem,...

> The intermediate slice b-spline registration is clever. Probably, but I think I overdid this. 62 into 270 (50um resolution) is to much for such a kind of an interpolation....

Because of the ongoing issues with the edge detection for the annotation file I created my own version. Instead of using a canny like edge detector I abused the fact...

> Would you be interested in opening a PR with the code you have so far, and we can work on it together there? We could add it to the...