COVE: COntext and VEracity prediction for out-of-context images