We track points sampled on the first frame. Please note that only CoTracker and RealTracker can track through occlusions.
However, CoTracker loses tracked points at the end while we are still tracking them.
|
BootsTAPIR
|
LocoTrack
|
CoTracker
|
Ours offline
|
We track 10k points sampled on a regular grid starting from the initial video frame.
Since the points are grid-sampled, tracks without significant transformations should maintain grid patterns in future frames.
LocoTrack and RealTracker tracks are better aligned than BootsTAPIR tracks. Neither LocoTrack nor BootsTAPIR can track through occlusions.
They also lose more background and object points than RealTracker.
|
BootsTAPIR
|
LocoTrack
|
Ours offline
|
Scaling helps improve both models, while in these examples the online model benefits from scaling more than the offline one.
|
RealTracker online base (Ours)
|
RealTracker online scaled (Ours)
|
RealTracker offline base (Ours)
|
RealTracker offline scaled (Ours)
|