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
Modern tracking algorithms use DNN internally (heavy task for JS and browser).
Customer feedback about intelligent scissors:
Fully finished polygons often have two points in one place
Sometimes there are strong deviations of automatically created line which are difficult to control because even the slightest cursor shift creates different lines
It is necessary to add an option of "automatic bordering" for existing polygons. It would help in tasks where you need to allocate all objects in the frame without empty areas.
Sometimes when new points are created quickly, the following message is displayed: RuntimeError: float unrepresentable in integer range
8111568 - Exception catching is disabled, this exception cannot be caught. Compile with -s DISABLE_EXCEPTION_CATCHING=0 or DISABLE_EXCEPTION_CATCHING=2 to catch.
The cursor and the square are poorly visible
It is necessary to add a limit to the possible size of the threshold or a warning about possible slowdown of the working process (when the action threshold is very big the performance is low).
Due to the square shape of the action's threshold, it turns out that diagonal lines can be longer than horizontal and vertical ones. Perhaps, it would be better to develop an option to rotate the square or change the action's threshold shape to a circle instead of the square.
The text was updated successfully, but these errors were encountered:
My actions before raising this issue
The nearest steps:
LUV or YUV should work too with equalizeHist() (less computations)
Also possible features for later milestones:
OpenCV contrib is not well supported yet.
Has very limited approach (detects bounding box/boxes of similar objects). Scaling or rotation are not supported (from the box).
May help if we have many similar objects ( something like that: https://www.microscopeworld.com/p-3468-microscope-resolution-explained-using-blood-cells.aspx ). But we still need multiple patterns.
Modern tracking algorithms use DNN internally (heavy task for JS and browser).
Customer feedback about intelligent scissors:
RuntimeError: float unrepresentable in integer range
8111568 - Exception catching is disabled, this exception cannot be caught. Compile with -s DISABLE_EXCEPTION_CATCHING=0 or DISABLE_EXCEPTION_CATCHING=2 to catch.
The text was updated successfully, but these errors were encountered: