superGMM_UCM_avg_sLuv + BIC

Post date: Sep 20, 2011 8:10:33 PM

Experiment#1

[url]: file:///home/student1/MATLABcodes/Collection_Algorithms_results/BSDS300_data/superGMM_superKmean/BENCH_UCM_superGMM_superKmean/color/bnd_superGMM_ucm_avg_sLuv_norm_multBnd/main.html

BDE

The Pb of this method when contours are summed together: F = 0.68

Experiment#2: Using BIC to choose the optimal number of cluster by varying the value of alpha. When using only 1% of the pixels.

Pick the best alpha of this experiment, then go to the next experiment on the number of pixels to use.

Experiment#3: Using BIC to choose the optimal number of cluster by varying the percent of the pixel. When using the best alpha from the previous experiment.

Pick the best alpha of this experiment, then go to the next experiment on the number of pixels to use.