Pure Noise to the Rescue of Insufficient Data: Improving Imbalanced Classification by Training on Random Noise Images

Open paper
Benchmark Model Rank Results
long-tail-learning-on-cifar-10-lt-r-100OPeN (WideResNet-28-10)#8Error Rate: 13.9
long-tail-learning-on-cifar-10-lt-r-50OPeN (WideResNet-28-10)#4Error Rate: 10.8
long-tail-learning-on-cifar-100-lt-r-100OPeN (WideResNet-28-10)#12Error Rate: 45.8
long-tail-learning-on-cifar-100-lt-r-50OPeN (WideResNet-28-10)#10Error Rate: 40.2
long-tail-learning-on-imagenet-ltOPeN (ResNeXt-50)#37Top-1 Accuracy: 55.1
long-tail-learning-on-places-ltOPeN (ResNet-152)#15Top-1 Accuracy: 40.5

Rank counts only results with a code link.