Table 2.

Best Rand and information theoretic scores (before and after border thinning) of all teams and the human experts using the undisclosed test set as of November 4, 2013.

Method	VRand	VInfo	VRand (thinned)	VInfo (thinned)
Human 1 vs. consensus	0.997 ± 0.001	0.997 ± 0.001	0.998 ± 0.001	0.999 ± 0.001
Human 2 vs. consensus	0.971 ± 0.003	0.941 ± 0.002	0.990 ± 0.002	0.989 ± 0.001
IDSIA-SCI	0.979 ± 0.005	0.988 ± 0.002	0.979 ± 0.005	0.988 ± 0.002
IDSIA-optree	0.969 ± 0.006	0.977 ± 0.003	0.972 ± 0.006	0.984 ± 0.002
SCI	0.966 ± 0.006	0.984 ± 0.002	0.968 ± 0.006	0.984 ± 0.002
IDSIA	0.944 ± 0.011	0.969 ± 0.002	0.978 ± 0.004	0.988 ± 0.001
BlackEagles	0.930 ± 0.009	0.941 ± 0.003	0.973 ± 0.005	0.983 ± 0.002
MLL-ETH	0.927 ± 0.008	0.926 ± 0.003	0.968 ± 0.006	0.981 ± 0.002
SDU	0.909 ± 0.011	0.926 ± 0.004	0.942 ± 0.008	0.974 ± 0.003
CellProfiler	0.904 ± 0.015	0.937 ± 0.006	0.915 ± 0.015	0.958 ± 0.005
Coxlab	0.901 ± 0.012	0.936 ± 0.006	0.939 ± 0.012	0.976 ± 0.003
Harvard	0.892 ± 0.017	0.944 ± 0.006	0.907 ± 0.016	0.957 ± 0.003
CoMPLEX	0.877 ± 0.019	0.903 ± 0.008	0.890 ± 0.018	0.947 ± 0.005
MLA	0.875 ± 0.016	0.885 ± 0.004	0.916 ± 0.016	0.964 ± 0.004
ML	0.867 ± 0.016	0.879 ± 0.006	0.911 ± 0.016	0.958 ± 0.003
UCL	0.860 ± 0.020	0.939 ± 0.005	0.863 ± 0.020	0.948 ± 0.005
TSC+PP	0.843 ± 0.012	0.839 ± 0.006	0.922 ± 0.013	0.961 ± 0.005
CLP	0.839 ± 0.024	0.885 ± 0.008	0.869 ± 0.024	0.940 ± 0.006
IMMI	0.826 ± 0.022	0.862 ± 0.008	0.876 ± 0.020	0.948 ± 0.005
ICOS	0.809 ± 0.018	0.838 ± 0.011	0.883 ± 0.015	0.936 ± 0.004
Freiburg	0.800 ± 0.026	0.839 ± 0.007	0.835 ± 0.027	0.928 ± 0.006
NIST	0.730 ± 0.021	0.757 ± 0.007	0.796 ± 0.020	0.851 ± 0.006
Computer Vision Jena	0.709 ± 0.024	0.768 ± 0.012	0.832 ± 0.022	0.904 ± 0.007
Bar-Ilan	0.701 ± 0.034	0.792 ± 0.011	0.773 ± 0.032	0.872 ± 0.012

For each team, the submission with the highest score is chosen for each column. The values were computed as the mean and standard error over the n = 20 test images that were not used in the public leaderboard.