How to implement an unknown class in neural nets?

I am currently engaged in a deep learning project for ‘multi-class insects classification’. I am working with 10 classes of insects. I want to define an unknown class in the model i.e. 10+1. In other words, when an input image comes in for the prediction, which doesn’t belong to any of the 10 classes, then the model should predict it as an unknown class.

I am unable to define this unknown class in my model. If anyone have any idea for this, please help me.

Thanx in advance :slight_smile: