OpenMax Explorer

What is OpenMax?

OpenMax is a method for detecting out-of-distribution (OOD) samples in deep learning models, based on the idea that the final layer activation values of OOD samples also out-of-distribution. OpenMax calculates the OpenMax score for each class, which is the probability that the input is in the class, and the probability that the input is OOD. The OpenMax probability is used to determine whether the input is in-distribution or OOD. If the OOD probability is above a certain threshold ε, the input is classified as OOD. ε is a hyperparameter that can be tuned to adjust the sensitivity of the model to OOD samples. For this classifier based on ResNet-56, ε is good at 0.05.