3D insert

There is a video file in which the girl turns her head.

You need to put a 3D object on the girl’s head.

How to calculate the camera to do this.

For example, I use two splines to calculate the camera. The first spline is on the wall. The second spline is created on the girl’s moving head.

When calculating the camera CameraSolve, the accuracy is 38 percent. The “Auto” mode is set

How to increase the accuracy of the camera calculation?

What is the camera doing in this shot?

If the camera is only panning, tilting, and zooming, then you can use the single back plane to calculate the camera motion and create a camera solve.

If there’s parallax, you don’t have enough non-coplanar planes to solve this shot.

You never including animating objects in a camera solve, only stationary ones. Once you solve for the camera, any animated objects like her head will export with the camera data when you select them. But you can’t include them in the solve or the solve will be inaccurate.

Please let me know if you have any more questions and I will be happy to help you.

I need to insert a 3D object on the girl’s head.

And for this you need to improve the accuracy of the camera calculation.

Where in this example do you need to place the splines to improve the accuracy of the calculation?

You can’t use her head to solve the shot. Only stationary objects. Once finished, you can export the head tracking data WITH the solve. But you can’t solve with her head because it is moving.

What is the camera doing? Is it on a tripod or handheld?

The camera is on a tripod.

OK, then use the single BG wall track to calculate the camera solve. Then select the tracked head layer and export the camera data. You will have a tracking dataset that moves with the head.

I did so and got “Solve Quality 38%”. Then I selected both layers in Mocha Pro and export the data through CameraSolve. But After Effects creates 5 layers of null object per spline, and there are no origin points in each null layer.

If the solve isn’t above 90 with a pan tilt zoom model, I’d not use it. Try getting a better track on the back wall.

  1. Created some splines in the background and raised Solve Quality from 38 to 83%

  2. Performed the export of data into After Effects

  3. After inserting data into After Effects, a series of layers of null objects are created that are not visible in the composition.

A video of my actions and the result obtained is attached

It may be that this shot is not a candidate for the camera solver. 83% isn’t really a result I think I would want to use.