Updates based on Beat's feedback

Author: stekunda

art/attacks/evasion/adversarial_patch/adversarial_patch_pytorch.py

@@ -490,7 +490,8 @@ def generate(  # type: ignore
 
         :param x: An array with the original input images of shape NCHW or input videos of shape NFCHW.
         :param y: True or target labels of format `list[dict[str, Union[np.ndarray, torch.Tensor]]]`, one for each
-                  input image. For untargeted attacks, this should be None. The fields of the dict are as follows:
+                  input image. For untargeted attacks, if `y` is None the attack will attack the model predictions
+                  on `x`. The fields of the dict are as follows:
 
                   - boxes [N, 4]: the boxes in [x1, y1, x2, y2] format, with 0 <= x1 < x2 <= W and 0 <= y1 < y2 <= H.
                   - labels [N]: the labels for each image.

art/attacks/evasion/dpatch_robust.py

@@ -155,9 +155,7 @@ def generate(  # type: ignore
         if y is None and self.targeted:
             raise ValueError("The targeted version of RobustDPatch attack requires target labels provided to `y`.")
         if y is not None and not self.targeted:
-            raise ValueError(
-                "The untargeted version of RobustDPatch attack does not use target labels provided to 'y'."
-            )
+            raise ValueError("The untargeted version of RobustDPatch attack does not use True labels provided to 'y'.")
         if x.ndim != 4:  # pragma: no cover
             raise ValueError("The adversarial patch can only be applied to images.")