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
I recently updated from an older (0.11.5) to newer (3.2.7) version of multiview-reconstruction and noticed a pretty significant change in behavior in how interest points are displayed in the interactive detection mode for difference of gaussian. I've attached images showing the difference I see in the old vs. new implementation. After playing around, it seems like I ultimately get very similar numbers of interest points when the sigma and threshold are set to the same values, but only a (random?) subset of the detections seem to be displayed in the newer implementation. Is this the expected behavior? It makes it more difficult to assess when the parameters have been set to desired values.
Thanks in advance for the help!
The text was updated successfully, but these errors were encountered:
AFAIK this interacitve DoG issue has been around since commit 855ed15 or shortly after (early October 2022), immediately following multiview-reconstruction-0.11.5, when net.preibisch.legacy.segmentation.InteractiveDoG was abandoned for imglib2 based code.
Since the issue doesn't seem to affect the ultimate detections, I think it has gone unnoticed.
@StephanPreibisch this issue should be improved in #55. However, the behavior of interactive DoG is somewhat a matter of personal taste, and the pull request alters the plugin's behavior and may not please everyone.
Hi,
I recently updated from an older (0.11.5) to newer (3.2.7) version of multiview-reconstruction and noticed a pretty significant change in behavior in how interest points are displayed in the interactive detection mode for difference of gaussian. I've attached images showing the difference I see in the old vs. new implementation. After playing around, it seems like I ultimately get very similar numbers of interest points when the sigma and threshold are set to the same values, but only a (random?) subset of the detections seem to be displayed in the newer implementation. Is this the expected behavior? It makes it more difficult to assess when the parameters have been set to desired values.
Thanks in advance for the help!
The text was updated successfully, but these errors were encountered: