detectMoreLoopClosures CLI using a lot of RAM in comparison to databaseViewer

[matlabbe]

The reason is that in the CLI tool we use Rtabmap object to use the internal Rtabmap::detectMoreLoopClosures function. However, by doing so, it indirectly loads the whole visual vocabulary and all descriptors in RAM. Would need to check how to do the same without Rtabmap object, or having an option to create Rtabmap object without loading all visual descriptors on initialization.

rtabmap/tools/DetectMoreLoopClosures/main.cpp

Lines 222 to 232 in 802b8f9

	Rtabmap rtabmap;
	printf("Initialization...\n");
	uInsert(parameters, inputParams);
	rtabmap.init(parameters, dbPath);
	float xMin, yMin, cellSize;
	bool haveOptimizedMap = !rtabmap.getMemory()->load2DMap(xMin, yMin, cellSize).empty();
	PrintProgressState progress;
	printf("Detecting...\n");
	int detected = rtabmap.detectMoreLoopClosures(clusterRadiusMax, clusterAngle, iterations, intraSession, interSession, &progress, clusterRadiusMin);

Original post: http://official-rtab-map-forum.206.s1.nabble.com/Matching-during-reprocessing-tp11148p11166.html