User Manual

Using Seed Frames
The solver works by first constructing a partial solution between two seed frames. These seed
frames are selected automatically. However, automatic selection adds time to the process. The
time is reported in the solve summary at the top of the Inspector once you run the solver. You
can select your own seed frames to speed the process and potentially get a better solve on
trickier clips. The solver uses these seed frames to create an intermediate solution and then
extends this forward and backward for the duration of the clip.
Selecting appropriate seed frames is not necessarily recommended unless you have some
experience with camera tracking. Keeping the default Auto Select Seed Frames checkbox
enabled in the Solve Options section of the Solver tab selects the best frames in most cases.
However, you can disable the checkbox and use the Seed Frame 1 and Seed Frame 2 slider to
select frames you believe achieve better results.
When choosing seed frames, it is important to satisfy two main characteristics:
 Seed frames should have lots of tracks in common.
 Seed frames should be far apart in perspective.
Sometimes There’s Nothing You Can Do
Some shots that do not have enough camera motion to triangulate feature locations cannot
be reconstructed with any useful accuracy. Ensuring that a shot is camera-trackable begins
on set, with proper usage of track markers, and by ensuring that camera moves have
enough perspective shifts for the solver to glean useful data from.
Cleaning Up Camera Solves
Sometimes the first solve will be good enough. Other times, it may take a few hours of cleaning
up tracks to get a good solve, and sometimes it is impossible. With experience, one gets a feel
for which tracks should be deleted and which should be kept, and which shots will be easy,
difficult, or impossible to solve.
Be aware that deleting too many tracks can cause the Average Solve Error to increase, as the
solver has too little information to work with. In particular, if there are fewer than eight tracks on
any frame, mathematically there is not enough information to solve the clip. However, it is
strongly recommended to use a lot more than eight tracks to get a robust and accurate solve.
IMPORTANT: If you are not familiar with camera tracking, it may be tempting to try to
directly edit the resulting 3D splines in the Spline Editor in order to improve a solved
camera’s motion path. This option should be used as an absolute last resort. It’s
preferable, instead, to modify the 2D tracks being fed into the solver.
1598Chapter – 77 3D Camera Tracking