How confident is a classification neural network for images with its prediction?

Hi all,

I trained a neural network to classify images. I trained it to identify circles and squares in the image. Now I want it to classify an image of a triangle. I didn’t use triangles in my training data. So the neural network isn’t familiar with it, and will classify it as either circle or square. However, since the network doesn’t know triangles, it somehow shouldn’t be confident with its classification. Is there a way, to get an information on how confident the neural network is with its prediction?

Thanks a lot for your help!

Sven