You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The merge mode in neuroglancer does not limit merges to be made with selected supervoxels. A mis-click, especially with small processes, can cause a merge into a random unselected neighboring object. This is particularly dangerous because the m key to initiate merge mode is right next to the keys for moving between sections and the activation of merge mode is extremely subtle given its dramatic impact (#471).
Undo will help fix the outcomes of this, but I think would still be proper to require that merges be between selected objects. While one can imagine that imagery continuity is good enough and there may be an argument for allowing this, having to load the object on the other side requires that the proofreader at least look at the segmentation as a preview before doing a merge, which can avoid merging good segmentations into bad highly-erroneous segmentations and I think is good practice anyway.
The text was updated successfully, but these errors were encountered:
a related change would be to have 'merge mode' provide the user some visual feedback that they are in merge mode... changing the border of the window, or the cursor type, really anything as some misclicks can be because people forgot or accidentally entered merge mode (easy to do when the section buttons <>, are right next to m).
The merge mode in neuroglancer does not limit merges to be made with selected supervoxels. A mis-click, especially with small processes, can cause a merge into a random unselected neighboring object. This is particularly dangerous because the
m
key to initiate merge mode is right next to the keys for moving between sections and the activation of merge mode is extremely subtle given its dramatic impact (#471).Undo will help fix the outcomes of this, but I think would still be proper to require that merges be between selected objects. While one can imagine that imagery continuity is good enough and there may be an argument for allowing this, having to load the object on the other side requires that the proofreader at least look at the segmentation as a preview before doing a merge, which can avoid merging good segmentations into bad highly-erroneous segmentations and I think is good practice anyway.
The text was updated successfully, but these errors were encountered: