DoubleMatch: Improving Semi-Supervised Learning with Self-Supervision

Benchmark Model Rank Results
semi-supervised-image-classification-on-stl-10-1000-labelsDoubleMatch#1Accuracy: 95.65±0.20
semi-supervised-image-classification-on-svhn-1000-labelsDoubleMatch#2Accuracy: 97.90 ± 0.07
semi-supervised-image-classification-on-cifar-100-10000-labelsDoubleMatch#3Percentage error: 21.22± 0.17
semi-supervised-image-classification-on-svhn-250-labelsDoubleMatch#3Accuracy: 97.63±0.35
semi-supervised-image-classification-on-svhn-40-labelsDoubleMatch#5Percentage error: 15.37±11.81
semi-supervised-image-classification-on-cifar-100-2500-labelsDoubleMatch#8Percentage error: 27.07± 0.26
semi-supervised-image-classification-on-cifar-10-250-labelsDoubleMatch#10Percentage error: 5.56±0.42
semi-supervised-image-classification-on-cifar-10-4000-labelsDoubleMatch#11Percentage error: 4.65±0.17
semi-supervised-image-classification-on-cifar-100-400-labelsDoubleMatch#11Percentage error: 41.83± 1.22
semi-supervised-image-classification-on-cifar-10-40-labelsDoubleMatch#14Percentage error: 13.59±5.60