Changes between Version 57 and Version 58 of NTTData

Timestamp:

Oct 7, 2021, 2:01:02 PM (4 years ago)

Author:

Elaine Meng

Comment:

--

NTTData

@@ -369 +369 @@
 - 5ire virus generation, seems OK now. 
 - What colors to use for 7+ chains? Phil: can we have a palette repeat?  Eric, yes among the other possibilities. Phil likes the interpolated one but wants to check with Darrell and get back to us. Tom G suggests doing something fancier based on symmetries, maybe as a longer-term goal. Phil: another fancy coloring is by domain.
-- What to do when map level produces disconnected surface bits, with only a very small bit left after hide dust? Elaine: Doesn't that mean the level is wrong?  Tom: may be correct level for step 1 but gets fragmented at the higher step automatically used for larger data. Even a single piece may be too fragile if it's a thin shell. Phil: printing software automatically thickens a too-thin shell. More discussion of just using step 1 (too many vertices is problematic for printing), map smoothing, iterating threshold adjustment vs. "measure volume" with "includeMasked false" (may be too slow).  May need to leave this one to manual interactive adjustment. Tom G has fancier ideas like struts, or binning (e.g. bin 4 instead of step 4, more like smoothing, but requires also lowering threshold to avoid more fragmentation). Phil: measure volume of largest bit at step 1 vs 4, then possibly bin.
+- What to do when map level produces disconnected surface bits, with only a very small bit left after hide dust? Elaine: Doesn't that mean the level is wrong?  Tom: may be correct level for step 1 but gets fragmented at the higher step automatically used for larger data. Even a single piece may be too fragile if it's a thin shell. Phil: printing software automatically thickens a too-thin shell. More discussion of just using step 1 (too many vertices is problematic for printing), map smoothing, iterating threshold adjustment vs. "measure volume" with "includeMasked false" (may be too slow).  May need to leave this one to manual interactive adjustment. Chimera/X initial default is to put 1% of voxels inside contour. Tom G has fancier ideas like struts, or binning (e.g. bin 4 instead of step 4, more like smoothing, but requires also lowering threshold to avoid more fragmentation). Phil: measure volume of largest bit at step 1 vs 4, then possibly bin. Maybe enhance bin to use max instead of avg.