From: Efficient video collection association using geometry-aware Bag-of-Iconics representations
Training dataset | Video | Frame | Time (h) | |||||
---|---|---|---|---|---|---|---|---|
 | Total | Registered | Total | Keyframe | Registered | Iconics | Keyframe | Clustering |
Augmented Videoscape | 901 | 659 | 605,544 | 2784 | 2575 | 622 | 0.05 | 0.11 |
London YouTube | 15,417 | 4691 | 184,189,894 | 4,518,792 | 1,402,825 | 97,048 | 179.58 | 4.68 |
Berlin YouTube | 13,984 | 4317 | 179,380,784 | 3,395,501 | 1,086,560 | 83,262 | 165.47 | 4.30 |