Hey Adobe, any chance of getting the same result from the camera tracker twice?

[Aviram84]

Say I 3d tracked a footage, but I want to extend that footage a couple of more frame. Camera tracker starts the analysis again, I expect everything to stay the same, but no. The results are completely different. only after about 4-5 tries I got back the results I wanted. Why is that? Maybe I'm doing something wrong?

thanks

[brian704]

It will depend on the features it tracks. You can put masks on the pre-comped footage to restrict the area it looks at to calculate the camera. That may help to get a better track. The idea of the tracker is not to return the same solve, but to get the best solve.

[Mylenium]

Maybe I'm doing something wrong?

Sure. You are using AE. 'nuff said. AE's 3D tracker is garbage for these kinds of things since it doesn't allow calibration or manual track markers. That#s al lthere is to it. If you want replicable results, use SynthEyes, mocha Pro and so on.

Mylenium

[Rick Gerard · Accepted solution]

To get the same results first place a marker on the layer where you set the origin and ground plane, and added your first element, then take a screenshot so you can remember the points selected, then extend the out point, then run camera tracker again, then move to the marker and pick the same points you originally chose and set an origin and ground plane, move to where you added your first 3D element and add it again using the same tracking points, then you should have the same results. Even Syntheyes and Mocha pro may give you different results if you re-track a scene with different in and out points. You have got to use the exact same features to add the camera every time unless you have perfectly corrected any lens distortion in the shot and have a 100% perfect camera solution.

An alternate workflow if the placed 3D object has already left the frame would be to split the layer, add a new section of footage, run the camera tracker again and add a new camera. You then cut the camera layers to match the cuts in the footage. I have recently been doing this a lot with drone footage for real estate developers. I stick things on the ground that are going to only be in frame for four or five seconds, then need to add more things. I just cut the shot up into pieces and camera track as many times as I need to to get the shot done. It saves a ton of time and works just fine.