Is there a function directly for 3d reconstruction?

[lemonci]

If I have already known the camera poses (from a simulator) of the images, is there anyway for 3d reconstruction directly from glomap (a function equal to colmap point_triangulator )? I tried the latter one but the result was not so good. Thanks.

[StonerLing]

Actually GLOMAP use the triangulation function of COLMAP:

glomap/glomap/controllers/track_retriangulation.cc

Lines 13 to 24 in 18a70ee

	bool RetriangulateTracks(const TriangulatorOptions& options,
	const colmap::Database& database,
	std::unordered_map<camera_t, Camera>& cameras,
	std::unordered_map<image_t, Image>& images,
	std::unordered_map<track_t, Track>& tracks) {
	// Following code adapted from COLMAP
	auto database_cache =
	colmap::DatabaseCache::Create(database,
	options.min_num_matches,
	false, // ignore_watermarks
	{} // reconstruct all possible images
	);

And in my opinion, precise camera poses and keypoint tracks are enough for colmap point_triangulator.