Fig. 5From: Efficient video collection association using geometry-aware Bag-of-Iconics representationsExamples of identified landmark groups. Best view in colorBack to article page