IoU layer size fixed to batch size

[aditya1709]

Why is the IoU layer associated with a batch size? Is there a way to make it independent of the batch size?
It prevents us from analyzing the learned model without supplying the same number of images as the batch size.