dense: detect and improve tower like structure reconstruction #1017
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
add debug for why not tower
# Conflicts: # libs/MVS/Camera.h
|
Here is an example of tower-mode 3:
|
Merged
cdcseacave
added a commit
that referenced
this pull request
Jul 4, 2023
- add python API - COLMAP support in MvgMvsPipeline.py - interface for binary COLMAP - interface for Polycam scenes - tower mode #1017 - estimate 3D points scale - transform scene by a given transform matrix - unify Docker scripts and add support for GUI - fix incorrect comparison in BufferedOutputStream #1010 - add lines structure - compute the focus of attention of a set of cameras - add image mask support in mesh texturing
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Depth-map estimation for tower like structures suffer from bad neighbor views selection. This is caused by the fact that traditionally SfM stage recovers many more points in the background compared to point on the tower structure, the region of interest. Combined with the fact that the background is much farther away compared to the foreground surface, results in the neighbor selection being biased towards good background surface coverage, and poor or next to none overlap for the foreground surface.
This in turn is caused by several factors, but two are the most important:
The solution proposed here to overcome this problem is to detect if the structure to be reconstructed is tower like, in which case to approximate the tower coverage in 3D and sample it uniformly into a virtual sparse point cloud, which will be used for neighbor selection instead of the original point cloud.
Buy default this mode is enabled, but the tower like structure detection is very strict in deciding if it is indeed a tower. For now only scenes with up direction on positive Z are supported, with units in meters.