Post-solve Checks

< Previous | Contents | Manuals Home | Boris FX | Next >

Post-solve Checks

After completing any solve, you should check the results to verify that they are physically plausible. Here are some checks to perform (if they are included in your solve).

Verify that the C*nn parameters are under ±0.5, decreasing in absolute value as you consider second-order, fourth-order, and sixth-order

terms. Physically, these parameter values are the maximum change they cause at the corners of the image. Though different values can offset each other, multiple large offsetting values typically indicate a touchy solve. Note that values may be substantially larger after running the Lens Workflows script; this is not an issue as long as the original values were OK.

Verify that the lens center coordinate HCTR, VCTR are less than 0.1 or maybe 0.2. Larger values indicate severely skewed solves; values over 1 indicate that the lens center is completely off screen.

Verify that any HSCL or VSCL values are close to 1. HSCL should never be solved unless field of view is set to Known. VSCL values should probably stay between 0.8-1.2.

Rotation angle should be less than 2 degrees. You shouldn’t manually create any animation on it, as it represents a static physical orientation.

Rolling shutter fraction should be less than one (the entire frame) and greater than zero, unless the shot was filmed inverted. Typical values should be in the 0.2-0.9 range.

Anamorphic distance can only be interpreted once a physical coordinate system has been set up (meters, feet, cm, etc). The

anamorphic distance is in that same coordinate system, so it should be at most roughly comparable to the length of the camera, including lens (due to the complexity of the lenses, this isn’t a direct physical requirement). It can be positive or negative, depending on the lens.

Click on the Lens Room on the top tab bar, then look at shape of the lens grid in the camera view, including location of lens center. Be alert

for localized waviness in any of the lines. You can think about the lens grid as indicating how any straight edges in either the real world or CG renders will be mapped into the image. Similarly, you can compare the grid lines to any actual straight lines in the imagery.

Keep some zero-weighted trackers (ZWTs) to use as canaries. The canaries should be distributed through the shot and stay visible for as long as possible. As you refine your solve, you must be alert for

situations where the solved error goes down, but the error of your ZWT canaries goes up. When that happens, it indicates overfitting, that the solver is trying to hard, and has too little contradictory information, so that it adheres excessively and unphysically to the tracking data, becoming worse in the process than simpler solves.

If you have issues as described in the list above, you should examine the tracking data and solve in more detail. Errors might indicate bad trackers, bad links from trackers to a lidar mesh, focus breathing (possibly including animated distortion), or the need to include additional factors such as rolling shutter or anamorphic distance.


©2023 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.