iFRAME (inhomogeneous Filters Random Field And Maximum Entropy)
Experiment 2: Learning Sparse FRAME by Generative Boosting with Local Normalization
Code
Case 1: Cat
(1) Training images
(2) Comparison of local normalization using different window sizes (scaleFactor = 0.5, 1, 2, 3, and 4)
Synthesis images
movie 1 | movie 2 | movie 3 | movie 4 | movie 5
Sketch templates
(3) Parameters setting:
General Parameters: nOrient = 16; sizeTemplatex=100;
sizeTemplatey=100; GaborScaleList=[ 1.4, 1, 0.7, 0.5]; DoGScaleList =[]; LocationShiftLimit=2; OrientShiftLimit=1; interval=5; #Wavelet=300; nIteration= (interval) x (#Wavelet) x 2;
Local normalization Parameters: isLocalNormalize=true; isSeparate=false; thresholdFactor=0.01; localNormScaleFactor=0.5, 1, 2, 3, or 4;
Gibbs Sampler Parameters: lambdaLearningRate = 0.1/sqrt(sigsq);
6x6 chains; sigsq=10; threshold_corrBB=0; lower_bound_rand = 0.001; upper_bound_rand = 0.999; c_val_list=-30:2:30
Case 2: Lion
(1) Training images
(2) Comparison of local normalization using different window sizes (scaleFactor = 0.5, 1, 2, 3, and 4)
Synthesis images
movie 1 | movie 2 | movie 3 | movie 4 | movie 5
Sketch templates
(3) Parameters setting:
General Parameters: nOrient = 16; sizeTemplatex=100;
sizeTemplatey=100; GaborScaleList=[ 1.4, 1, 0.7, 0.5]; DoGScaleList =[]; LocationShiftLimit=2; OrientShiftLimit=1; interval=5; #Wavelet=300; nIteration= (interval) x (#Wavelet) x 2;
Local normalization Parameters: isLocalNormalize=true; isSeparate=false; thresholdFactor=0.01; localNormScaleFactor=0.5, 1, 2, 3, or 4;
Gibbs Sampler Parameters: lambdaLearningRate = 0.1/sqrt(sigsq);
6x6 chains; sigsq=10; threshold_corrBB=0; lower_bound_rand = 0.001; upper_bound_rand = 0.999; c_val_list=-30:2:30
Case 3: Hummingbird
(1) Training images
(2) Comparison of local normalization using different window sizes (scaleFactor = 0.5, 1, 2, 3, and 4)
Synthesis images
movie 1 | movie 2 | movie 3 | movie 4 | movie 5
Sketch templates
(3) Parameters setting:
General Parameters: nOrient = 16; sizeTemplatex=100;
sizeTemplatey=100; GaborScaleList=[ 1.4, 1, 0.7, 0.5]; DoGScaleList =[]; LocationShiftLimit=2; OrientShiftLimit=1; interval=5; #Wavelet=300; nIteration= (interval) x (#Wavelet) x 2;
Local normalization Parameters: isLocalNormalize=true; isSeparate=false; thresholdFactor=0.01; localNormScaleFactor=0.5, 1, 2, 3, or 4;
Gibbs Sampler Parameters: lambdaLearningRate = 0.1/sqrt(sigsq);
6x6 chains; sigsq=10; threshold_corrBB=0; lower_bound_rand = 0.001; upper_bound_rand = 0.999; c_val_list=-30:2:30
Case 4: Tiger
(1) Training images
(2) Comparison of local normalization using different window sizes (scaleFactor = 0.5, 1, 2, 3, and 4)
Synthesis images
movie 1 | movie 2 | movie 3 | movie 4 | movie 5
Sketch templates
(3) Parameters setting:
General Parameters: nOrient = 16; sizeTemplatex=100;
sizeTemplatey=100; GaborScaleList=[ 1.4, 1, 0.7, 0.5]; DoGScaleList =[]; LocationShiftLimit=2; OrientShiftLimit=1; interval=5; #Wavelet=300; nIteration= (interval) x (#Wavelet) x 2;
Local normalization Parameters: isLocalNormalize=true; isSeparate=false; thresholdFactor=0.01; localNormScaleFactor=0.5, 1, 2, 3, or 4;
Gibbs Sampler Parameters: lambdaLearningRate = 0.1/sqrt(sigsq);
6x6 chains; sigsq=10; threshold_corrBB=0; lower_bound_rand = 0.001; upper_bound_rand = 0.999; c_val_list=-30:2:30
