Mish: A Self Regularized Non-Monotonic Activation Function

Open paper
Benchmark Model Rank Results
image-classification-on-cifar-10ResNet 9 + Mish#158Percentage correct: 94.05
image-classification-on-cifar-10ResNet v2-20 (Mish activation)#177Percentage correct: 92.02
image-classification-on-cifar-100ResNet v2-110 (Mish activation)#146Percentage correct: 74.41
image-classification-on-imagenetCSPResNeXt-50 + Mish#698Top 1 Accuracy: 79.8%

Rank counts only results with a code link.