Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Appearance settings

Switch to perform clustering on all registered frames #4007

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lpanaf merged 4 commits into from
Jan 15, 2026
Merged

Switch to perform clustering on all registered frames #4007

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ca8b94f
switch to perform pruning on all registered frames
lpanaf Jan 14, 2026
f3c127f
remove step numbers
lpanaf Jan 14, 2026
85caad0
Merge remote-tracking branch 'origin/main' into lpanaf/prune_all_regi…
lpanaf Jan 15, 2026
74c9e6b
d
lpanaf Jan 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 5 additions & 33 deletions 38 src/colmap/scene/reconstruction_clustering.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,7 +171,7 @@ std::unordered_map<frame_t, int> ClusterReconstructionFrames(
}
}

// Step 2: Filter edges to keep only reliable connections.
// Filter edges to keep only reliable connections.
std::unordered_map<image_pair_t, int> edge_weights;
for (const auto& [pair_id, count] : frame_covisibility_count) {
if (count < options.min_covisibility_count) continue;
Expand All @@ -185,36 +185,8 @@ std::unordered_map<frame_t, int> ClusterReconstructionFrames(
return {};
}

// Step 3: Keep only the largest connected component and de-register the rest.
std::vector<std::pair<frame_t, frame_t>> edges;
edges.reserve(edge_weights.size());
for (const auto& [pair_id, weight] : edge_weights) {
const auto [frame_id1, frame_id2] = PairIdToImagePair(pair_id);
edges.emplace_back(frame_id1, frame_id2);
}
const std::vector<frame_t> largest_cc_vec =
FindLargestConnectedComponent(nodes, edges);
const std::unordered_set<frame_t> largest_cc(largest_cc_vec.begin(),
largest_cc_vec.end());
for (const auto& [frame_id, frame] : reconstruction.Frames()) {
if (largest_cc.count(frame_id) == 0 && frame.HasPose()) {
reconstruction.DeRegisterFrame(frame_id);
}
}
LOG(INFO) << "Kept " << largest_cc.size() << " frames in largest component";

// Filter to keep only edges within the largest component.
for (auto it = edge_weights.begin(); it != edge_weights.end();) {
const auto [frame_id1, frame_id2] = PairIdToImagePair(it->first);
if (largest_cc.count(frame_id1) == 0 || largest_cc.count(frame_id2) == 0) {
it = edge_weights.erase(it);
} else {
++it;
}
}

// Step 4: Compute adaptive threshold using median minus median absolute
// deviation (MAD). Extract weight values after filtering to largest CC.
// Compute adaptive threshold using median minus median absolute
// deviation (MAD).
std::vector<int> weight_values;
weight_values.reserve(edge_weights.size());
for (const auto& [pair_id, weight] : edge_weights) {
Expand All @@ -225,8 +197,8 @@ std::unordered_map<frame_t, int> ClusterReconstructionFrames(
std::max(median - mad, options.min_edge_weight_threshold);
LOG(INFO) << "Threshold for Strong Clustering: " << threshold;

// Step 5: Cluster frames based on covisibility weights.
Comment thread
lpanaf marked this conversation as resolved.
return EstablishStrongClusters(options, largest_cc, edge_weights, threshold);
// Cluster frames based on covisibility weights.
return EstablishStrongClusters(options, nodes, edge_weights, threshold);
}

} // namespace colmap
10 changes: 3 additions & 7 deletions 10 src/colmap/scene/reconstruction_clustering.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,21 +71,17 @@ struct ReconstructionClusteringOptions {
// Algorithm:
// 1. Build a covisibility graph where edges connect frames sharing >=
// min_covisibility_count points.
// 2. Find the largest connected component and de-register all frames outside
// it. This removes isolated or poorly connected parts of the
// reconstruction.
// 3. Compute an adaptive edge weight threshold using median minus median
// 2. Compute an adaptive edge weight threshold using median minus median
// absolute deviation (MAD).
// 4. Cluster frames using union-find: first merge strongly connected frames,
// 3. Cluster frames using union-find: first merge strongly connected frames,
// then iteratively merge clusters connected by multiple weaker edges.
//
// Args:
// options: Configuration options for clustering.
// reconstruction: The reconstruction containing frames and 3D points.
// Frames outside the largest connected component will be de-registered.
//
// Returns:
// Map from frame_id to cluster_id for frames in the largest component.
// Map from frame_id to cluster_id for all registered frames.
std::unordered_map<frame_t, int> ClusterReconstructionFrames(
const ReconstructionClusteringOptions& options,
Reconstruction& reconstruction);
Expand Down
Loading
Morty Proxy This is a proxified and sanitized view of the page, visit original site.