Spaces:
Runtime error
Runtime error
A note about the algorithm
#1
by
Puddisnky
- opened
This might sound odd, but i see how the algorithm works to do this, and it's elegant and decently effective and fast. Now at the risk of sounding daft, I think I know a better, albeit a similar way this can be done. For about the same time and calculations, I'm fairly sure you can avoid the 'pockets' of unchecked image, scale the sensitivity dynamically but better and add some additional accuracy also. I'm no coder though, so if someone would like details, I'm willing to share.
You are right but I didn't changed any params of the original model and weights.
asim266
changed discussion status to
closed