diff --git a/docs/source/en/tasks/object_detection.md b/docs/source/en/tasks/object_detection.md
index 39362b46158..dfad80b949f 100644
--- a/docs/source/en/tasks/object_detection.md
+++ b/docs/source/en/tasks/object_detection.md
@@ -204,6 +204,8 @@ Instantiate the image processor from the same checkpoint as the model you want t
 ```py
 >>> from transformers import AutoImageProcessor
 
+>>> MAX_SIZE = IMAGE_SIZE
+
 >>> image_processor = AutoImageProcessor.from_pretrained(
 ...     MODEL_NAME,
 ...     do_resize=True,
@@ -225,8 +227,6 @@ and it uses the exact same dataset as an example. Apply some geometric and color
 ```py
 >>> import albumentations as A
 
->>> max_size = IMAGE_SIZE
-
 >>> train_augment_and_transform = A.Compose(
 ...     [
 ...         A.Perspective(p=0.1),