Results for IJCNN 2011 competition (1st stage)
Please find below all results that were submitted during the first stage of the IJCNN 2011 competition "The German Traffic Sign Recognition Benchmark" between January 19, and January 21, 2011. The top contestants won a free registration for the conference and were invited to participate in final live competition session that was held at IJCNN 2011 in San Jose, CA, USA. The results of this final session are published in the official GTSRB result table.
Subset | |||
---|---|---|---|
[197] IDSIA | cnn_hog3 | 98.98% | 98.98% |
[196] IDSIA | cnn_cnn_hog3 | 98.98% | 98.98% |
[178] sermanet | EBLearn 2LConvNet ms 108 feats | 98.97% | 98.97% |
[195] IDSIA | cnn_cnn_hog3_haar | 98.97% | 98.97% |
[187] sermanet | EBLearn 2LConvNet ms 108 + val | 98.89% | 98.89% |
[199] INI-RTCV | Human performance | 98.81% | 98.81% |
[170] IDSIA | CNN(IMG)_MLP(HOG3) | 98.79% | 98.79% |
[177] IDSIA | CNN(IMG)_MLP(HOG3)_MLP(HAAR) | 98.72% | 98.72% |
[26] sermanet | EBLearn 2-layer ConvNet ms | 98.59% | 98.59% |
[193] IDSIA | CNN 7HL norm | 98.46% | 98.46% |
[185] sermanet | EBLearn 2L CNN ms + validation | 98.41% | 98.41% |
[198] sermanet | EBLearn 2-layer ConvNet ms reg | 98.41% | 98.41% |
[27] sermanet | EBLearn 2-layer ConvNet ss | 98.20% | 98.20% |
[191] IDSIA | CNN 7HL | 98.10% | 98.10% |
[183] Radu.Timofte@VISICS | IKSVM + PHOG + HOG2 | 97.88% | 97.88% |
[166] IDSIA | CNN 6HL | 97.56% | 97.56% |
[184] Radu.Timofte@VISICS | SRC + LDAs I/HOG1/HOG2 | 97.35% | 97.35% |
[176] Radu.Timofte@VISICS | CS+HOGs+limited | 97.01% | 97.01% |
[84] noob | HOG+LDA+VQ | 96.87% | 96.87% |
[157] noob | Combined HOG+VQ+LDA | 96.79% | 96.79% |
[175] italian_crash | MultiDataset Alg Fix 5 | 96.78% | 96.78% |
[73] noob | Subspace Analysis 9 | 96.74% | 96.74% |
[87] noob | Subspace Analysis 13 | 96.73% | 96.73% |
[70] noob | Subspace Analysis 7 | 96.67% | 96.67% |
[90] noob | Subspace Analysis 14 | 96.65% | 96.65% |
[98] Radu.Timofte@VISICS | CS+HOGs | 96.63% | 96.63% |
[77] noob | Subspace Analysis 11 | 96.61% | 96.61% |
[101] noob | HOG2 + VQ + Variant of LDA | 96.59% | 96.59% |
[146] Tomato | HOGNoScaleLDA | 96.53% | 96.53% |
[136] Tomato | HOG+LDA | 96.53% | 96.53% |
[125] matiko | LDA+4SVM | 96.45% | 96.45% |
[124] matiko | 2LDA+3SVM | 96.44% | 96.44% |
[194] matiko | LDA+OVA | 96.44% | 96.44% |
[129] matiko | 2LDA+6SVM, various features | 96.44% | 96.44% |
[127] matiko | 2LDA+5SVM | 96.42% | 96.42% |
[76] noob | Subspace Analysis 10 | 96.42% | 96.42% |
[179] shn | 2L NL SFA(HOG02)+RG+GC | 96.40% | 96.40% |
[123] matiko | 2LDA+2SVM | 96.40% | 96.40% |
[128] matiko | 2LDA+4SVM - various features | 96.40% | 96.40% |
[137] Tomato | HOG0203 + LDA | 96.40% | 96.40% |
[138] noob | HOG + Data Clustering | 96.39% | 96.39% |
[122] matiko | LDA+2SVM | 96.38% | 96.38% |
[132] matiko | 2LDA+2SVM (2) | 96.37% | 96.37% |
[94] masaki | whybrid1 | 96.34% | 96.34% |
[106] Tomato | HOG2 + matlab LDA | 96.32% | 96.32% |
[2] INI-RTCV | HOG features (Set 2) + LDA | 96.32% | 96.32% |
[68] masaki | mergef5 | 96.31% | 96.31% |
[162] nistor | HOG2 nomralize with LDA | 96.31% | 96.31% |
[89] masaki | 0002 | 96.30% | 96.30% |
[95] masaki | scale23s4c4 | 96.29% | 96.29% |
[83] masaki | mergef8 | 96.29% | 96.29% |
[82] masaki | scale23s4c_5 | 96.29% | 96.29% |
[78] masaki | scale23s4de | 96.29% | 96.29% |
[88] masaki | 0001 | 96.28% | 96.28% |
[93] masaki | hybrid3 | 96.28% | 96.28% |
[67] masaki | mergef4 | 96.28% | 96.28% |
[85] masaki | mergeff | 96.28% | 96.28% |
[86] masaki | 0000 | 96.27% | 96.27% |
[57] masaki | merge5 | 96.25% | 96.25% |
[81] masaki | mergef7 | 96.24% | 96.24% |
[55] masaki | mergef - mix, logis,23 cm2 | 96.20% | 96.20% |
[65] masaki | mergef3 | 96.20% | 96.20% |
[79] masaki | mergef6 | 96.20% | 96.20% |
[180] matiko | LDA+OVO+OVA | 96.16% | 96.16% |
[91] masaki | hybird-scale23s4de-mergef8 | 96.07% | 96.07% |
[54] masaki | mix23scale | 96.07% | 96.07% |
[92] masaki | hybrid2 | 96.07% | 96.07% |
[56] masaki | mergeall | 96.05% | 96.05% |
[168] titan | linearSVM_ECOC_ensemble | 95.90% | 95.90% |
[104] Tomato | HOG_02-L2RL2SVM | 95.89% | 95.89% |
[108] Tomato | HOG_02_CramerAndSinger | 95.85% | 95.85% |
[43] masaki | Ensemble | 95.83% | 95.83% |
[140] noob | HOG + Diaglinear Analysis | 95.82% | 95.82% |
[33] masaki | libsvm --s | 95.82% | 95.82% |
[53] masaki | logistic23 | 95.82% | 95.82% |
[45] masaki | Ensemble 4 | 95.82% | 95.82% |
[18] masaki | linear SVM param selection | 95.81% | 95.81% |
[50] masaki | mix23 | 95.78% | 95.78% |
[63] masaki | scale23vote-s0 | 95.73% | 95.73% |
[40] masaki | CMmodel | 95.72% | 95.72% |
[66] masaki | hog02s4c4w1 | 95.72% | 95.72% |
[80] masaki | HOG02s16c4w1 | 95.72% | 95.72% |
[46] masaki | cmmodel 2 | 95.72% | 95.72% |
[64] masaki | HOG02s4c2w1 | 95.72% | 95.72% |
[141] titan | linearSVM_ECOC_HOG02 | 95.70% | 95.70% |
[16] masaki | linear SVM fixed | 95.60% | 95.60% |
[17] masaki | linear SVM fixed 2 | 95.60% | 95.60% |
[19] Diego Peteiro-Barral | Optimal 1-layer ANN (2) | 95.57% | 95.57% |
[23] Diego Peteiro-Barral | Optimal 1-layer ANN (4) | 95.57% | 95.57% |
[52] masaki | HOG2 | 95.54% | 95.54% |
[156] italian_crash | MultiDataset Alg Fix3 | 95.48% | 95.48% |
[24] Diego Peteiro-Barral | Optimal 1-layer ANN (5) | 95.47% | 95.47% |
[44] masaki | zeroLinearSVM | 95.47% | 95.47% |
[15] Diego Peteiro-Barral | Optimal 1-layer ANN | 95.47% | 95.47% |
[71] noob | Subspace Analysis 8 | 95.47% | 95.47% |
[22] Diego Peteiro-Barral | Optimal 1-layer ANN (3) | 95.47% | 95.47% |
[188] italian_crash | MultiDataset Alg Final | 95.35% | 95.35% |
[149] italian_crash | MultiDataset Alg Fix1 | 95.34% | 95.34% |
[150] italian_crash | multiDataset Alg Fix2 | 95.28% | 95.28% |
[51] masaki | reduced2 svm | 95.24% | 95.24% |
[173] shn | 2L NL SFA(HOG2) CCC | 95.19% | 95.19% |
[190] shn | SFA(SFA(IM)+HOG) | 95.16% | 95.16% |
[182] shn | 2L NL SFA(IM+HOG02)+CC | 95.12% | 95.11% |
[151] olbustosa | RandomForest_hog_hue_200tree | 95.05% | 95.05% |
[69] noob | Subspace Analysis 6 | 94.84% | 94.84% |
[3] INI-RTCV | HOG features (Set 3) + LDA | 94.73% | 94.73% |
[142] TDC | CVOG+HOG+texture+ANN | 94.73% | 94.73% |
[154] TDC | hybrid feature + ANN | 94.73% | 94.73% |
[172] nistor | HOG3 normalization + LDA | 94.70% | 94.70% |
[48] Brainsignals | met1 | 94.61% | 94.61% |
[1] INI-RTCV | HOG features (Set 1) + LDA | 94.51% | 94.51% |
[161] testn | HOG1 features + LDA | 94.48% | 94.48% |
[148] olbustosa | RandomForest_Hog_Hue_100tree | 94.47% | 94.47% |
[186] soumith | Cascaded ConvNet 70 feat | 94.41% | 94.41% |
[155] Brainsignals | simple alignment plus htsa | 94.24% | 94.24% |
[130] olbustosa | RandomForest_HOG_100tree | 94.18% | 94.18% |
[144] titan | linearSVM_Allpairs_HOG123 | 94.16% | 94.16% |
[171] titan | linearSVM_Allpairs_ensemble | 94.12% | 94.12% |
[189] soumith | Cascaded ConvNet Mixed arch | 94.06% | 94.06% |
[118] CAOR | KdTree HOG9720 kNN5 iter 1000 | 93.72% | 93.72% |
[145] titan | linearSVM_Allpairs_HOG02 | 93.67% | 93.67% |
[147] titan | linearSVM_Allpairs_HOG02_2 | 93.67% | 93.67% |
[103] CAOR | KdTree + HOG 9720 kNN 5 | 93.40% | 93.40% |
[102] CAOR | KdTree + HOG 9720 | 93.21% | 93.21% |
[100] CAOR | KdTree + HOG 2520 | 92.90% | 92.90% |
[174] CAOR | Random trees HOG3 | 92.13% | 92.13% |
[61] noob | Subspace Analysis 4 | 92.09% | 92.09% |
[143] TDC | HOG+textures+ANN | 91.99% | 91.99% |
[60] noob | Subspace Analysis 3 | 91.98% | 91.98% |
[139] Tomato | HOGs+LDA | 91.88% | 91.88% |
[105] TDC | CVOG+HOG(set1)+ANN | 91.81% | 91.81% |
[62] noob | Subspace Analysis 5 | 91.57% | 91.57% |
[121] olbustosa | RandomForest_Hog_20tree | 91.34% | 91.34% |
[49] masaki | Ensemble23 | 91.25% | 91.25% |
[159] italian_crash | MultiDataset Alg Fix4 | 91.16% | 91.16% |
[160] italian_crash | MultiDataset Alg Fix4.1 | 91.03% | 91.03% |
[99] TDC | HOG(set 1) + ANN | 90.99% | 90.99% |
[37] masaki | linear svm --s param selection | 90.39% | 90.39% |
[158] soumith | Cascaded cscscf ConvNet | 90.19% | 90.19% |
[34] shn | Second Network CC | 89.70% | 89.70% |
[164] MadRose | Hopfield Network | 88.84% | 88.84% |
[58] noob | Subspace Analysis 1 | 88.54% | 88.54% |
[119] CAOR | KdTree + HOG 3456 kNN 5 | 88.27% | 88.27% |
[10] noob | Nearest Subspace | 87.70% | 87.70% |
[21] shn | test CC | 87.02% | 87.02% |
[20] shn | test GC | 86.62% | 86.62% |
[75] noob | Nearest Subspace 2 | 86.18% | 86.18% |
[36] AlcalaGRAM2 | Contrast Gaussian param modif | 85.87% | 85.87% |
[30] Alcalá-GRAM | Contrast-Gaussian SVM | 84.96% | 84.96% |
[114] olbustosa | RandomForest-Hog | 83.24% | 83.24% |
[72] italian_crash | MultiDataset Algorithm | 83.08% | 83.08% |
[12] TDC | CVOG + CCV + NN (Team 2) | 82.67% | 82.67% |
[11] TDC | CVOG + CCV + NN (Team 1) | 82.37% | 82.37% |
[74] TDC | CVOG + ANN (Team 3) | 81.80% | 81.80% |
[97] RMULG | Subwindows+ETGRAY+LIBLINEAR | 79.71% | 79.71% |
[134] olbustosa | HOG_SVM | 76.35% | 76.35% |
[163] RMULG | Subwindows+Filters+ET+LIBLINEA | 76.31% | 76.31% |
[32] AlcalaGRAM2 | Gaussian SVM param selection | 75.67% | 75.67% |
[181] FiVENiNEs | Multiclass SVM with CV preproc | 74.83% | 74.83% |
[42] RMULG | Subwindows+ETGRAY | 74.80% | 74.80% |
[107] olbustosa | NaiveBayes | 74.64% | 74.64% |
[133] olbustosa | NaiveBayesMultinomial | 73.94% | 73.94% |
[7] INI-RTCV | HOG 1 + 3-NN (Euclidean) | 73.89% | 73.89% |
[9] INI-RTCV | HOG 3 + 3-NN (Euclidean) | 73.82% | 73.82% |
[6] INI-RTCV | HOG 3 + 1-NN (Euclidean) | 73.82% | 73.82% |
[4] INI-RTCV | HOG 1 + 1-NN (Euclidean) | 73.65% | 73.65% |
[5] INI-RTCV | HOG 2 + 1-NN (Euclidean) | 72.81% | 72.81% |
[8] INI-RTCV | HOG 2 + 3-NN (Euclidean) | 72.81% | 72.81% |
[59] noob | Subspace Analysis 2 | 72.71% | 72.71% |
[115] CAOR | KdTree HOG Set3 2 | 70.86% | 70.86% |
[109] CAOR | KdTree HOG Set1 | 70.65% | 70.65% |
[116] CAOR | KdTree HOG 2 2 | 68.80% | 68.80% |
[31] RMULG | Subwindows+ET | 67.43% | 67.43% |
[47] olbustosa | RandomForest_ini | 64.15% | 64.15% |
[41] RMULG | Subwindows+ETK28 | 59.64% | 59.64% |
[153] TDC | Voting+16ANN | 48.25% | 48.25% |
[169] bilbo | ANN~88/700s(sect. hue+moments) | 42.91% | 42.91% |
[152] bilbo | ANN~80/600 (sect. hue+moments) | 38.06% | 38.06% |
[165] bilbo | ANN~88/1200(sect. hue+moments) | 28.63% | 28.63% |
[13] noob | Nearest Subspace 1 | 18.77% | 18.77% |
[117] CAOR | KdTree Hue | 17.53% | 17.53% |
[167] nistor | HueHist normalization and LDA | 14.50% | 14.50% |
[14] masaki | linear SVM | 3.67% | 3.67% |
[35] nistor | SOM CL | 3.55% | 3.55% |
[96] nistor | sMsup | 2.32% | 2.32% |
[39] nistor | HOGNEW2 | 2.04% | 2.04% |
[38] nistor | HOGNEW | 2.04% | 2.04% |
[29] nistor | SOMCL | 1.93% | 1.93% |
[135] nistor | HOG1 LDA modified | 0.29% | 0.29% |
[192] IDSIA | CNN 7HL newnorm | 0.00% | 0.02% |
References
[1] HOG features (Set 1) + LDA, INI-RTCV
[2] HOG features (Set 2) + LDA, INI-RTCV
[3] HOG features (Set 3) + LDA, INI-RTCV
[4] HOG 1 + 1-NN (Euclidean), INI-RTCV
[5] HOG 2 + 1-NN (Euclidean), INI-RTCV
[6] HOG 3 + 1-NN (Euclidean), INI-RTCV
[7] HOG 1 + 3-NN (Euclidean), INI-RTCV
[8] HOG 2 + 3-NN (Euclidean), INI-RTCV
[9] HOG 3 + 3-NN (Euclidean), INI-RTCV
[10] Nearest Subspace, noob
[11] CVOG + CCV + NN (Team 1), TDC
[12] CVOG + CCV + NN (Team 2), TDC
[13] Nearest Subspace 1, noob
[14] linear SVM, masaki
[15] Optimal 1-layer ANN, Diego Peteiro-Barral
[16] linear SVM fixed, masaki
[17] linear SVM fixed 2, masaki
[18] linear SVM param selection, masaki
[19] Optimal 1-layer ANN (2), Diego Peteiro-Barral
[20] test GC, shn
[21] test CC, shn
[22] Optimal 1-layer ANN (3), Diego Peteiro-Barral
[23] Optimal 1-layer ANN (4), Diego Peteiro-Barral
[24] Optimal 1-layer ANN (5), Diego Peteiro-Barral
[26] EBLearn 2-layer ConvNet ms, sermanet
[27] EBLearn 2-layer ConvNet ss, sermanet
[29] SOMCL, nistor
[30] Contrast-Gaussian SVM, Alcalá-GRAM
[31] Subwindows+ET, RMULG
[32] Gaussian SVM param selection, AlcalaGRAM2
[33] libsvm --s, masaki
[34] Second Network CC, shn
[35] SOM CL, nistor
[36] Contrast Gaussian param modif, AlcalaGRAM2
[37] linear svm --s param selection, masaki
[38] HOGNEW, nistor
[39] HOGNEW2, nistor
[40] CMmodel, masaki
[41] Subwindows+ETK28, RMULG
[42] Subwindows+ETGRAY, RMULG
[43] Ensemble, masaki
[44] zeroLinearSVM, masaki
[45] Ensemble 4, masaki
[46] cmmodel 2, masaki
[47] RandomForest_ini, olbustosa
[48] met1, Brainsignals
[49] Ensemble23, masaki
[50] mix23, masaki
[51] reduced2 svm, masaki
[52] HOG2, masaki
[53] logistic23, masaki
[54] mix23scale, masaki
[55] mergef - mix, logis,23 cm2, masaki
[56] mergeall, masaki
[57] merge5, masaki
[58] Subspace Analysis 1, noob
[59] Subspace Analysis 2, noob
[60] Subspace Analysis 3, noob
[61] Subspace Analysis 4, noob
[62] Subspace Analysis 5, noob
[63] scale23vote-s0, masaki
[64] HOG02s4c2w1, masaki
[65] mergef3, masaki
[66] hog02s4c4w1, masaki
[67] mergef4, masaki
[68] mergef5, masaki
[69] Subspace Analysis 6, noob
[70] Subspace Analysis 7, noob
[71] Subspace Analysis 8, noob
[72] MultiDataset Algorithm, italian_crash
[73] Subspace Analysis 9, noob
[74] CVOG + ANN (Team 3), TDC
[75] Nearest Subspace 2, noob
[76] Subspace Analysis 10, noob
[77] Subspace Analysis 11, noob
[78] scale23s4de, masaki
[79] mergef6, masaki
[80] HOG02s16c4w1, masaki
[81] mergef7, masaki
[82] scale23s4c_5, masaki
[83] mergef8, masaki
[84] HOG+LDA+VQ, noob
[85] mergeff, masaki
[86] 0000, masaki
[87] Subspace Analysis 13, noob
[88] 0001, masaki
[89] 0002, masaki
[90] Subspace Analysis 14, noob
[91] hybird-scale23s4de-mergef8, masaki
[92] hybrid2, masaki
[93] hybrid3, masaki
[94] whybrid1, masaki
[95] scale23s4c4, masaki
[96] sMsup, nistor
[97] Subwindows+ETGRAY+LIBLINEAR, RMULG
[98] CS+HOGs, Radu.Timofte@VISICS
[99] HOG(set 1) + ANN, TDC
[100] KdTree + HOG 2520, CAOR
[101] HOG2 + VQ + Variant of LDA, noob
[102] KdTree + HOG 9720, CAOR
[103] KdTree + HOG 9720 kNN 5, CAOR
[104] HOG_02-L2RL2SVM, Tomato
[105] CVOG+HOG(set1)+ANN, TDC
[106] HOG2 + matlab LDA, Tomato
[107] NaiveBayes, olbustosa
[108] HOG_02_CramerAndSinger, Tomato
[109] KdTree HOG Set1, CAOR
[114] RandomForest-Hog, olbustosa
[115] KdTree HOG Set3 2, CAOR
[116] KdTree HOG 2 2, CAOR
[117] KdTree Hue, CAOR
[118] KdTree HOG9720 kNN5 iter 1000, CAOR
[119] KdTree + HOG 3456 kNN 5, CAOR
[121] RandomForest_Hog_20tree, olbustosa
[122] LDA+2SVM, matiko
[123] 2LDA+2SVM, matiko
[124] 2LDA+3SVM, matiko
[125] LDA+4SVM, matiko
[127] 2LDA+5SVM, matiko
[128] 2LDA+4SVM - various features, matiko
[129] 2LDA+6SVM, various features, matiko
[130] RandomForest_HOG_100tree, olbustosa
[132] 2LDA+2SVM (2), matiko
[133] NaiveBayesMultinomial, olbustosa
[134] HOG_SVM, olbustosa
[135] HOG1 LDA modified, nistor
[136] HOG+LDA, Tomato
[137] HOG0203 + LDA, Tomato
[138] HOG + Data Clustering, noob
[139] HOGs+LDA, Tomato
[140] HOG + Diaglinear Analysis, noob
[141] linearSVM_ECOC_HOG02, titan
[142] CVOG+HOG+texture+ANN, TDC
[143] HOG+textures+ANN, TDC
[144] linearSVM_Allpairs_HOG123, titan
[145] linearSVM_Allpairs_HOG02, titan
[146] HOGNoScaleLDA, Tomato
[147] linearSVM_Allpairs_HOG02_2, titan
[148] RandomForest_Hog_Hue_100tree, olbustosa
[149] MultiDataset Alg Fix1, italian_crash
[150] multiDataset Alg Fix2, italian_crash
[151] RandomForest_hog_hue_200tree, olbustosa
[152] ANN~80/600 (sect. hue+moments), bilbo
[153] Voting+16ANN, TDC
[154] hybrid feature + ANN, TDC
[155] simple alignment plus htsa, Brainsignals
[156] MultiDataset Alg Fix3, italian_crash
[157] Combined HOG+VQ+LDA, noob
[158] Cascaded cscscf ConvNet, soumith
[159] MultiDataset Alg Fix4, italian_crash
[160] MultiDataset Alg Fix4.1, italian_crash
[161] HOG1 features + LDA, testn
[162] HOG2 nomralize with LDA, nistor
[163] Subwindows+Filters+ET+LIBLINEA, RMULG
[164] Hopfield Network, MadRose
[165] ANN~88/1200(sect. hue+moments), bilbo
[166] CNN 6HL, IDSIA
[167] HueHist normalization and LDA, nistor
[168] linearSVM_ECOC_ensemble, titan
[169] ANN~88/700s(sect. hue+moments), bilbo
[170] CNN(IMG)_MLP(HOG3), IDSIA
[171] linearSVM_Allpairs_ensemble, titan
[172] HOG3 normalization + LDA , nistor
[173] 2L NL SFA(HOG2) CCC, shn
[174] Random trees HOG3, CAOR
[175] MultiDataset Alg Fix 5, italian_crash
[176] CS+HOGs+limited, Radu.Timofte@VISICS
[177] CNN(IMG)_MLP(HOG3)_MLP(HAAR), IDSIA
[178] EBLearn 2LConvNet ms 108 feats, sermanet
[179] 2L NL SFA(HOG02)+RG+GC, shn
[180] LDA+OVO+OVA, matiko
[181] Multiclass SVM with CV preproc, FiVENiNEs
[182] 2L NL SFA(IM+HOG02)+CC, shn
[183] IKSVM + PHOG + HOG2, Radu.Timofte@VISICS
[184] SRC + LDAs I/HOG1/HOG2, Radu.Timofte@VISICS
[185] EBLearn 2L CNN ms + validation, sermanet
[186] Cascaded ConvNet 70 feat, soumith
[187] EBLearn 2LConvNet ms 108 + val, sermanet
[188] MultiDataset Alg Final, italian_crash
[189] Cascaded ConvNet Mixed arch, soumith
[190] SFA(SFA(IM)+HOG), shn
[191] CNN 7HL, IDSIA
[192] CNN 7HL newnorm, IDSIA
[193] CNN 7HL norm, IDSIA
[194] LDA+OVA, matiko
[195] cnn_cnn_hog3_haar, IDSIA
[196] cnn_cnn_hog3, IDSIA
[197] cnn_hog3, IDSIA
[198] EBLearn 2-layer ConvNet ms reg, sermanet
[199] Human performance, INI-RTCV