Reputation: 11
Why does YOLO add 'background' class to the object detection even when we have not added that class? How to remove it?
I am trying to train YOLO v8x model on custom data. In the confusion matrix I saw that background class has been added which I haven't annotated. Later I observed this happening with YOLO v5 as well. If anyone has any idea on this , could you please tell me? Also, how to remove this class? and where can we check which all classes are considered in the model for training?
Upvotes: 1
Views: 4036
Reputation: 1323
You can access the model's classes list by calling model.names
.
It will not contain the 'background' class because this is the default concept defining false-positive predictions of your model when it sees non-existing objects in the background. No need to try to remove it. It is useful to show it in a confusion matrix, so you can estimate the model performance relative to if it tends to generate false-positive predictions and which of your classes are subject to this mistake.
Upvotes: 0