We should evaluate a strategy to completely remove people and cars from the texture reconstruction to get a better looking result.
We could start on evaluating an inpaint strategy (https://stackoverflow.com/a/76224570) but we could consider also other approaches
Result of the investigation should include:
- evaluate the feasibility using python libraries (e.g. opencv bindings). Let's consider people and cars separately.
- snippet of code for a proposed solution to run in the point_cloud_to_mesh stage applying this changes to undistorted images
This issue is partially connected to #60