Abstract | ||
---|---|---|
•A keyframe summary of a video can be viewed as a reference set for a 1-nn classifier.•We propose a Greedy Tabu Search (GTS) algorithm to find one keyframe per event.•Egocentric videos benefit from GTS more than third-person videos do.•Comparisons with user ground truth rank GTS higher than an excellent baseline.•CNN was the best among 7 low-level and high-level features spaces. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1016/j.jvcir.2018.02.010 | Journal of Visual Communication and Image Representation |
Keywords | Field | DocType |
Keyframe summary,Nearest neighbour classifier,Instance selection,Egocentric video,Feature representations | Computer vision,Nearest neighbour,Feature vector,Pattern recognition,Convolutional neural network,Ground truth,Artificial intelligence,Classifier (linguistics),Mathematics | Journal |
Volume | ISSN | Citations |
52 | 1047-3203 | 3 |
PageRank | References | Authors |
0.40 | 20 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Ludmila I. Kuncheva | 1 | 4 | 2.13 |
Paria Yousefi | 2 | 3 | 1.75 |
Jurandy Almeida | 3 | 431 | 35.15 |