Back to Algorithm Ranking.
[Grayscale] Normalized Cut Benchmark: Algorithm "Graclus"
Graclus segmentation algorithm
Graclus is a graph clustering algorithm for normalized cut (and other
weighted graph clustering objectives) that uses weighted
kernel k-means and multilevel methods. The algorithm is based on a
mathematical equivalence between weighted graph clustering objectives and the
weighted kernel k-means objective function, which shows how to construct
kernel matrices and select weights given a graph in order to monotonically
improve the graph clustering objective during each iteration of the weighted
kernel k-means algorithm. In order to make the algorithm as fast as possible
and to avoid poor local optima, Graclus uses a multilevel approach, which
applies the weighted kernel k-means algorithm at various levels of coarseness
of the input graph.
Plot of F-measure for boundary detection task.
Click on an image for additional details.
#1 (119082) Ncut=0.665
|
|
#2 (170057) Ncut=2.590
|
|
#3 (58060) Ncut=0.046
|
|
#4 (163085) Ncut=0.487
|
|
#5 (42049) Ncut=0.522
|
|
#6 (167062) Ncut=0.191
|
|
#7 (157055) Ncut=0.523
|
|
#8 (295087) Ncut=0.203
|
|
#9 (24077) Ncut=0.320
|
|
#10 (78004) Ncut=0.590
|
|
#11 (220075) Ncut=0.203
|
|
#12 (45096) Ncut=0.560
|
|
#13 (38092) Ncut=1.457
|
|
#14 (43074) Ncut=0.282
|
|
#15 (16077) Ncut=0.245
|
|
#16 (86000) Ncut=0.913
|
|
#17 (101085) Ncut=0.804
|
|
#18 (219090) Ncut=0.238
|
|
#19 (89072) Ncut=0.158
|
|
#20 (300091) Ncut=0.352
|
|
#21 (126007) Ncut=0.662
|
|
#22 (156065) Ncut=0.262
|
|
#23 (76053) Ncut=0.217
|
|
#24 (296007) Ncut=0.236
|
|
#25 (175032) Ncut=0.094
|
|
#26 (253027) Ncut=1.060
|
|
#27 (304034) Ncut=0.015
|
|
#28 (86016) Ncut=0.665
|
|
#29 (103070) Ncut=0.161
|
|
#30 (8023) Ncut=0.175
|
|
#31 (260058) Ncut=0.054
|
|
#32 (41033) Ncut=0.538
|
|
#33 (291000) Ncut=0.187
|
|
#34 (109053) Ncut=0.830
|
|
#35 (130026) Ncut=0.043
|
|
#36 (241004) Ncut=0.473
|
|
#37 (108082) Ncut=0.116
|
|
#38 (285079) Ncut=0.666
|
|
#39 (147091) Ncut=1.221
|
|
#40 (69040) Ncut=0.203
|
|
#41 (14037) Ncut=0.320
|
|
#42 (54082) Ncut=0.429
|
|
#43 (189080) Ncut=0.321
|
|
#44 (229036) Ncut=0.862
|
|
#45 (62096) Ncut=0.127
|
|
#46 (271035) Ncut=1.518
|
|
#47 (167083) Ncut=0.046
|
|
#48 (12084) Ncut=1.142
|
|
#49 (69015) Ncut=1.489
|
|
#50 (148089) Ncut=1.022
|
|
#51 (160068) Ncut=0.289
|
|
#52 (145086) Ncut=0.477
|
|
#53 (216081) Ncut=3.751
|
|
#54 (97033) Ncut=0.253
|
|
#55 (182053) Ncut=0.690
|
|
#56 (208001) Ncut=0.922
|
|
#57 (19021) Ncut=0.623
|
|
#58 (227092) Ncut=1.807
|
|
#59 (134035) Ncut=0.230
|
|
#60 (223061) Ncut=0.021
|
|
#61 (253055) Ncut=1.089
|
|
#62 (148026) Ncut=0.254
|
|
#63 (210088) Ncut=3.259
|
|
#64 (86068) Ncut=0.094
|
|
#65 (3096) Ncut=0.172
|
|
#66 (41069) Ncut=0.099
|
|
#67 (21077) Ncut=0.133
|
|
#68 (196073) Ncut=0.220
|
|
#69 (108070) Ncut=0.210
|
|
#70 (123074) Ncut=0.116
|
|
#71 (376043) Ncut=0.251
|
|
#72 (306005) Ncut=0.244
|
|
#73 (38082) Ncut=0.223
|
|
#74 (33039) Ncut=0.484
|
|
#75 (108005) Ncut=0.107
|
|
#76 (106024) Ncut=0.176
|
|
#77 (302008) Ncut=0.315
|
|
#78 (102061) Ncut=0.186
|
|
#79 (197017) Ncut=0.282
|
|
#80 (299086) Ncut=0.286
|
|
#81 (37073) Ncut=1.026
|
|
#82 (241048) Ncut=0.156
|
|
#83 (65033) Ncut=1.652
|
|
#84 (55073) Ncut=0.340
|
|
#85 (66053) Ncut=0.410
|
|
#86 (143090) Ncut=0.135
|
|
#87 (85048) Ncut=0.508
|
|
#88 (42012) Ncut=0.236
|
|
#89 (351093) Ncut=0.463
|
|
#90 (361010) Ncut=1.164
|
|
#91 (175043) Ncut=0.027
|
|
#92 (87046) Ncut=0.036
|
|
#93 (105025) Ncut=0.192
|
|
#94 (236037) Ncut=0.071
|
|
#95 (101087) Ncut=1.192
|
|
#96 (304074) Ncut=0.228
|
|
#97 (296059) Ncut=0.196
|
|
#98 (159008) Ncut=0.072
|
|
#99 (385039) Ncut=0.646
|
|
#100 (69020) Ncut=0.149
|
|
Page generated on 13-Oct-2008 16:25:04.