0
$\begingroup$

Before reading: I have already followed this popular post and this post on lowering solve error intensively, along with this guide and this guide.

However, when it comes to eliminating tracks that are problematic using the filter option in the clean up menu. Eliminating only the tracks that have a 3.0 and above and short lifespans, the solve error skyrockets.

What was a solve error of 2.4406 is now 207.2394

Before Image of a house with multiple tracking markers, of which 4 are selected After image of a house with multiple tracking markers, where the previous 4 markers have been deleted, and the solve error as risen to 200+

What can I do to fix this?

Is this a bug?

Edit: Here's another post with the same issue and no answers.

$\endgroup$
5
  • $\begingroup$ Is there enough parallax information in your camera movement. Have you identified the smartest keyframe and do they contain enough shared tracks? Are the tracks on enough depths? (I can see that this is not the case, you didn't use the postbox, use it!) $\endgroup$
    – Leander
    Commented Sep 5, 2019 at 14:52
  • $\begingroup$ If you have 10 tracks on a plane and 2 at different depths, then the depth track will likely have the highest solve error, but will also be the most inportant track to recreate the scene. The solution is not to remove them but to add additional tracks at useful geometric position so that focal length and f1 can be better optimized resulting a lower solve error. $\endgroup$
    – Leander
    Commented Sep 5, 2019 at 14:54
  • $\begingroup$ The video moves from one side of the house to another so I'm sure there is enough parallax going on for the camera solver to grasp (roughly 80 deg-ish). The post box is in frame for about 150 frames so I can't use that either. I would put tracks onto the house itself but the day I filmed it, clouds overhead, and a high ISO caused a relatively noisy with no contrast so I resulted to things that I could see and use upfront. $\endgroup$ Commented Sep 5, 2019 at 15:01
  • $\begingroup$ Frame 541 and 644 seem too close for a 1200 frame track. Set up the keyframes to have the most possible depth information and share the most keyframes. $\endgroup$
    – Leander
    Commented Sep 5, 2019 at 15:03
  • $\begingroup$ This was a great suggestion! I brought the solve error back down to 0.26 which is a pretty great number for such a shoddy clip. Though I am not sure if this is the solution to this specific issue?(or if the tracking will be that good) I will try it with another clip of which I had the same issue with and post an update. $\endgroup$ Commented Sep 5, 2019 at 15:11

0

You must log in to answer this question.

Browse other questions tagged .