Write OpenSfM parameters presets for different types of data.

[paulinus]

• basic
• large drone datasets (gps based matching, local bundle + global)
• linear datasets (order based matching, local bundle)
• high accuracy (more features, stricter thresholds, larger images and depthmaps)

[dakotabenjamin]

Currently we could do just local bundle or just global, but not the hybrid type that you mentioned in the call, right?

[paulinus]

We can do hybrid also. For example, run local bundle for every image added, then a global bundle from time to time. For example, every 100 images and everytime the number of points increases by 20% since last full bundle.

bundle_interval: 100            # Bundle after adding 'bundle_interval' cameras
bundle_new_points_ratio: 1.2    # Bundle when (new points) / (bundled points) > bundle_new_points_ratio
local_bundle_radius: 1          # Max image graph distance for images to be included in local bundle adjustment

[dakotabenjamin]

Oh that is perfect.

[dakotabenjamin]

FWIW I tested the hybrid bundle adjustment method on >5000 image dataset and that part worked. It crashes on merging depthmaps though, as I expected.

EDIT: I added 120GB more RAM and it worked fine: