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
Hi, good question. It is an artifact of the planner finding local minima, but there are a few tricks that sometimes help with this.
Try setting the --lr_multiplier to a larger value. Maybe 2.0 or so. This will help the strokes move around more and maybe avoid being placed in the background.
You can also increase the --init_optim_iter parameter to like 1000 or 2000 or more. This will give more optimization iterations to help the strokes move to better areas.
Thank you very much for your reply! I tried to modify the mentioned parameters and the effect was greatly improved. But as the init_optim_iter doubles, the drawing time will become very long. For example, in the case of RTX3090, num_strokes=100, and init_optim_iter=2000, it takes approximately 20 minutes to complete the painting. Do you have any inspiration to solve this problem? I will be very willing to try.
For speed, unfortunately the optimization process is pretty slow, and improving speed generally comes at the cost of the quality of the drawing. We don't have any implemented methods for greatly speeding up the process right now, unfortunately.
When I use Frida, there are many meaningless strokes on the background of the simulated image. May I ask how you handle these noises?
The text was updated successfully, but these errors were encountered: