hey folks,
I am extremely sorry if you found this question completely unrelated to this forum. But I am gonna ask anyway, why don't production house use camera that itself captures the motion data, instead of tracking the camera motion in post, why don't we record the camera motion on a shooting day itself ? Motion sensor doesn't cost much, and can be integrated easily into camera.

Hey swikarp,

good question! I asked this myself. I can only speculate but my theory is based on some experience: The motion tracker data is not precise enough and syncing it with the footage is cumbersome. One of my last projects was using a spike camera, an extremely precisely controlled robot camera. It exports its motion data natively. But even with this setup, the shots get tracked based on the captured footage. It's faster and more precise.

For a track, it must sit 100%. Even the smallest misalignment is recognizable. That's why I think it's the way it is.

Cheers
CYTE

oh, now I get it. Precision is the factor that's playing a huge role.
If someone makes a device preicse enough and easy to mount on camera, that would be revolutionary in the industry.
Maybe I should give it a try ( I wish I was a robotics engineer )
Anyway, thanks CYTE for the info !

I can also only speculate based on my understanding but It's not just the precision of the device, it's precision of the overall alignment with the scene

So even if the device itself gives you precise motion you still need to align it to 3D representation of the scene, whether it's lidar or gaussian splat or whatever you end up having

I dont know what type of motion sensors you are talking about, but for example tracking and accumulating just motion delta will give you motion relative to the start alignment at best and of course with the danger of losing accuracy as it accumulates over time

So you still need to either calibrate at least the start alignment to the 3d scene or of course every frame, either during capture with some external sensors or post capture, which is exactly what most matchmove software do

So in theory yes, if you live in an ideal world and can record all necessary data (motion, alignment, 3d scene representation) precisely it would probably give you perfect alignment, but in real world it's probably much easier to compute best fit alignment per frame and get visually as perfect representation of the camera motion within the imperfect 3d scene representation than having perfect relative motion that doesn't align to anything

You can probably always combine all the information you have, like sensor data, in tracking to get to the result easier especially for difficult to track shots, but then it may be a question of whether its worth capturing and syncing it if you'd still need to use image based alignment and match move techniques anyway

But again,just my limited understanding of the issues

Now I get the whole picture. Really Appreciate your effort Tomas.
Thanks