Fig. 7From: Efficient video collection association using geometry-aware Bag-of-Iconics representationsRetrieval pipeline overviewBack to article page