In mocha documentation there is one tip regarding rotoscoping:
While you can refine a shape you have tracked to do your rotoscoping, the recommended way is to do a rough shape to track something and then link your roto to that track.
I used this method to roto a passenger bus passing by and it worked very well. But I wonder if the same method is recommended to roto walking humans/animals?
I have a footage of a person walking (side view), so I tracked the person's head and then tried to link my roto shapes to the track. But the roto shapes of the limbs do not follow the track very well. And I noticed that it's difficult to track walking persons head (as I understand, if I want to link my roto shapes to a track, the track itself must be almost perfect).
So, should I use the method described above to roto objects like this? Maybe I should re-shoot the video with a person walking in the background with more contrast?
Here is the footage which I'm working with: https://www.youtube.com/watch?v=k0VuJT0Xu50
Note that I do not intend to rotoscope entire 3 seconds - I only need 13 frames (during which the walking person completes 2 steps)