From: Efficient video collection association using geometry-aware Bag-of-Iconics representations
Dataset | Number of images | Processing time (h) | |||||
---|---|---|---|---|---|---|---|
 | Registered | Iconics | Total | Stream | Densify | SfM | Total |
Berlin Flickr | 865,699 | 37,544 | 2,661,327 | 18.46 | 1.89 | 5.57 | 25.92 |
London Flickr | 3,716,916 | 103,290 | 12,036,991 | 90.75 | 7.09 | 33.83 | 131.67 |
 | Videos | Length (h) | Frames | Keyframes | Registered | Clusters |  |
London YouTube | 19,217 | 2,195.96 | 245,586,526 | 5,648,490 | 734,303 | 4,937 | Â |
Berlin YouTube | 17,480 | 2,068.41 | 223,388,274 | 4,244,377 | 636,689 | 4,135 | Â |