Contributed to the keras-team/keras repository by enhancing the extract_patches function to support configurable stride parameters, improving flexibility in patch-based image preprocessing. Addressed a bug in the strides handling logic, ensuring more accurate and consistent patch extraction for computer vision workflows. The work focused on enabling default stride values and reducing edge-case failures, which streamlines downstream data augmentation and preprocessing pipelines. Leveraged Python and image processing expertise, with a strong emphasis on testing to validate the new functionality. This update reduces maintenance overhead for users and allows models to operate with a wider range of stride configurations in their preprocessing steps.