The Prague Texture Segmentation Datagenerator and Benchmark - Result Details
optimized
for

It is possible that the resulting index maps don't correspond to the training set's indeces according to the data.xml file!
It leads to the wrong values of pixel-wise criteria as a result.


Benchmark - Colour [normal] CLresult byhuangyuan's unet (psp21ewt)
set \ criterionCSOSUSMENEOCCACOCCIIIEAMSRMCIGCELCEBCEGBCEBGMSCSSCVDLAVINVINMIMARIJCDCFMIWIWIINBDE
#3 - all99.780.000.000.000.000.210.1599.5699.7899.780.220.1299.7899.670.0699.780.440.350.530.4499.7899.5699.560.2299.780.271.5398.470.3099.3299.1099.5599.5599.5599.550.17
#4 - all98.930.000.000.000.000.941.1397.8998.9398.931.070.3798.9398.400.0998.932.121.982.262.1298.9397.8997.891.0798.861.094.8895.011.0697.2996.1098.0198.0197.9898.040.63
#5 - all99.350.000.000.000.000.650.4798.7299.3599.360.650.1999.3599.030.0799.351.281.141.421.2899.3598.7298.720.6599.140.702.7197.060.5098.6798.0499.0199.0198.9299.100.30
#6 - all98.230.000.000.000.002.051.8996.5198.2398.241.770.6998.2297.340.3098.233.442.644.303.4998.2396.5496.511.7797.651.585.5193.681.5395.8994.0096.9196.9196.2697.560.88
#7 - all97.590.000.000.000.003.272.3095.3997.5997.602.410.3497.5996.380.1997.594.654.205.124.6797.5995.3895.392.4196.212.076.6592.921.1995.8693.3796.5796.5796.7496.411.01
#8 - all98.960.000.000.000.001.120.9797.9598.9698.971.040.1598.9698.450.0798.962.051.822.292.0698.9697.9597.951.0498.801.063.1896.690.5297.8996.4698.2098.2098.2898.110.36
#9 - all96.2727.120.000.840.003.013.6693.1496.2795.813.730.4895.9994.400.5296.025.994.687.896.5796.2793.6493.223.3586.582.567.2691.581.7494.0490.6595.0995.1095.6194.592.14
#10 - all97.900.000.000.000.002.202.0395.9497.9097.932.100.2097.9196.860.2097.914.083.404.804.1297.9095.9195.942.1097.491.945.2594.600.8195.9493.0596.4096.4096.8795.930.63
#11 - all98.190.000.000.000.001.771.3396.4898.1998.211.810.1998.1997.290.1298.203.543.044.043.5498.1996.4796.481.8197.741.684.3795.410.6796.6994.3197.0797.0797.0997.050.52
#12 - all98.410.000.000.000.001.461.4196.8898.4198.431.590.1798.4097.610.1598.413.092.423.803.1298.4196.9096.881.5997.321.453.6495.970.6397.1195.0697.4797.4797.6697.280.45
#6 - bark98.360.000.000.000.001.371.9796.7998.3698.361.640.4098.3697.540.1798.363.202.763.663.2298.3696.8096.791.6497.831.515.2794.221.1796.5794.7997.3397.3397.2497.410.70
#6 - flowers97.770.000.000.000.002.552.9095.6997.7797.792.230.4397.7796.660.2897.784.343.734.954.3497.7795.6995.692.2397.131.966.8292.721.3695.8593.6296.7196.7196.7296.700.85
#6 - glass97.010.000.000.000.002.223.2894.2397.0197.002.990.7396.9995.510.2797.005.745.186.395.8397.0194.2994.232.9996.042.488.6590.712.0693.7090.4895.0095.0094.6695.351.56
#6 - man-made99.060.000.000.000.000.910.9198.1399.0699.060.940.2199.0698.580.1299.061.871.622.121.8899.0698.1398.130.9498.970.983.4396.360.7097.8396.6198.2898.2898.4098.150.45
#6 - nature99.240.000.000.000.000.870.8798.4999.2499.240.760.3999.2498.850.0999.241.511.291.731.5199.2498.4998.490.7698.780.802.7796.090.7998.3498.0098.9998.9999.0298.960.35
#6 - plants97.070.000.000.000.001.361.4994.3397.0797.132.930.8697.0595.600.8697.075.493.917.215.6397.0794.4194.332.9397.322.167.5092.252.2892.5288.5493.9293.9293.2994.561.42
#6 - rock98.950.000.000.000.001.131.1297.9298.9598.951.050.2498.9598.420.0798.952.091.912.272.0998.9597.9297.921.0598.821.053.6796.140.7797.5596.1398.0398.0398.0698.000.41
#6 - stone96.660.000.002.551.891.192.7797.1098.4698.551.540.2898.4397.700.4198.472.682.113.422.8698.4697.2097.101.5494.361.154.0294.440.7398.3997.9198.9498.9498.9398.950.69
#6 - textile99.470.000.000.000.000.600.5498.9699.4799.470.530.1299.4799.210.0399.471.050.981.121.0599.4798.9598.960.5399.270.602.0897.740.3798.8998.2799.1399.1399.1199.140.22
#6 - wood94.646.900.003.950.002.585.1490.3094.6491.485.361.3392.9891.961.7293.025.364.769.619.0194.6493.4390.854.4183.252.558.8889.642.3893.6190.8395.1995.2392.5697.981.69
mean98.091.700.000.370.091.571.8296.5298.1898.011.820.3998.0897.270.2998.093.202.703.953.4498.1896.7196.551.7596.571.484.9094.581.0896.6094.7797.2997.2997.1597.440.77
criterionGraph
correct
over
under
missed
noise
F-measure
imageTexture mosaic (undegraded)Ground truthSegmentation
#3 - all
#4 - all
#5 - all
#6 - all
#7 - all
#8 - all
#9 - all
#10 - all
#11 - all
#12 - all
#6 - bark
#6 - flowers
#6 - glass
#6 - man-made
#6 - nature
#6 - plants
#6 - rock
#6 - stone
#6 - textile
#6 - wood