Compute residuals with threadpool for model_merger #3401
Conversation
This file contains hidden or 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
ahojnnes
reviewed
Jun 12, 2025
src/colmap/estimators/alignment.cc
Outdated
| const Image& src_image = *src_images[i]; | ||
| const Image& tgt_image = *tgt_images[i]; | ||
| if (thread_pool_ && src_images.size() > 10) { | ||
| std::vector<std::future<void>> futures; |
Contributor
There was a problem hiding this comment.
No need to accumulate the futures. You can just do thread_pool_->Wait(); to finish all tasks in the pool.
Author
There was a problem hiding this comment.
Fixed but does not propagate THROW_CHECK_EQ exceptions. How could this be solved?
Author
There was a problem hiding this comment.
I removed THROW_CHECK_EQ from ComputeImageResidual, and only perform the checks once before computing the residuals.
ahojnnes
reviewed
Jun 12, 2025
src/colmap/sfm/observation_manager.h
Outdated
| #include "colmap/scene/image.h" | ||
| #include "colmap/scene/reconstruction.h" | ||
| #include "colmap/scene/track.h" | ||
| #include "colmap/util/threading.h" |
Contributor
There was a problem hiding this comment.
Let's forward-decl threadpool here as well instead of including this header.
Author
|
Hey @ahojnnes , |
sarlinpe
reviewed
Jul 11, 2025
src/colmap/estimators/alignment.cc
Outdated
| } else { | ||
| for (size_t i = 0; i < src_images.size(); ++i) { | ||
| ComputeImageResidual( | ||
| i, src_images, tgt_images, tgt_from_src, src_from_tgt, residuals); |
Member
There was a problem hiding this comment.
How about instead passing src_images[i], tgt_images[i], and residuals[i]?
Author
|
Hey @sarlinpe , is this okay to merge? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Hey @ahojnnes ,
We've recently had a computation bottleneck when trying to merge big reconstructions (2,500 to 35,000 common images). The heaviest step is computing the residuals for alignment. This PR simply computes the residuals with a threadpool.
On two models with 2,706 common images the alignment step drops from 646 s to 51 s with 64 threads.
N.B.:
THROW_CHECK_EQexceptions in threads are propagated in the main thread throughfuture.get().