Fix: Support 'jpg' format in keras.utils.save_img()

[utsab345]

Summary

This PR fixes issue #21678.
The save_img function previously failed with .jpg extension because Pillow only supports 'JPEG'.
Now, .jpg is automatically mapped to 'jpeg'.

Changes

• Updated keras/utils/image_utils.py to map 'jpg' → 'jpeg'.
• Added test test_save_img_with_jpg in integration_tests/test_save_img.py.

Notes

• Verified with pytest, all tests pass.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[gemini-code-assist]

Summary of Changes

Hello @utsab345, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a compatibility issue within the keras.utils.save_img utility, enabling it to properly handle images specified with the common '.jpg' extension. By internally converting 'jpg' to 'jpeg', the function now works seamlessly with underlying image processing libraries, enhancing the robustness of image saving operations within Keras.

Highlights

• Bug Fix: The keras.utils.save_img() function now correctly supports the '.jpg' file extension, resolving an issue where it previously failed due to Pillow's requirement for 'JPEG'.
• Format Mapping: The save_img function has been updated to automatically map 'jpg' to 'jpeg' internally, ensuring compatibility with image processing libraries like Pillow.
• New Tests: New integration tests (test_save_jpg_rgb and test_save_jpg_rgba) have been added to verify the correct saving of images with the '.jpg' extension for both RGB and RGBA formats.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses an issue with saving images in JPG format by mapping 'jpg' to 'jpeg', which is the format name Pillow expects. The change is a good step, but the logic for handling RGBA to RGB conversion for JPEGs is incomplete as it doesn't account for cases where the file format is inferred from the path. I've suggested a more robust implementation for this. Additionally, the new tests could be improved by using parametrization to reduce duplication and by adding more comprehensive assertions to verify the saved image content. My feedback includes suggestions to address these points.

[gemini-code-assist]

This change correctly handles file_format='jpg', but it misses the case where file_format is None and the path ends with .jpg. This will cause a crash when saving an RGBA image. The logic can be made more robust by also inspecting the path extension to determine if the output is a JPEG. Also, the condition on line 181 becomes partially redundant after converting jpg to jpeg.

Suggested change

	if file_format is not None and file_format.lower() == 'jpg':
	file_format = 'jpeg'
	img = array_to_img(x, data_format=data_format, scale=scale)
	if img.mode == "RGBA" and (file_format == "jpg" or file_format == "jpeg"):
	if img.mode == "RGBA" and file_format in ["jpeg", "jpg"]:
	_format = file_format
	if _format is None and isinstance(path, (str, pathlib.Path)):
	_format = pathlib.Path(path).suffix[1:].lower()
	if file_format is not None and file_format.lower() == 'jpg':
	file_format = 'jpeg'
	img = array_to_img(x, data_format=data_format, scale=scale)
	if img.mode == "RGBA" and _format in ("jpeg", "jpg"):

[gemini-code-assist]

These tests are a good start, but they can be improved. The two test functions are very similar and can be combined into a single parameterized test using pytest.mark.parametrize for better maintainability. Also, the assertions only check for file existence. It would be more robust to load the saved image and verify its content, especially to confirm that RGBA images are correctly converted to RGB.

import os

import numpy as np
import pytest
from keras.utils import img_to_array, load_img, save_img


@pytest.mark.parametrize(
    "shape, name",
    [
        ((50, 50, 3), "rgb.jpg"),
        ((50, 50, 4), "rgba.jpg"),
    ],
)
def test_save_jpg(tmp_path, shape, name):
    img = np.random.randint(0, 256, size=shape, dtype=np.uint8)
    path = tmp_path / name
    save_img(path, img, file_format="jpg")
    assert os.path.exists(path)

    # Check that the image was saved correctly and converted to RGB if needed.
    loaded_img = load_img(path)
    loaded_array = img_to_array(loaded_img)
    assert loaded_array.shape == (50, 50, 3)

[codecov-commenter]

Codecov Report

❌ Patch coverage is 80.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 76.87%. Comparing base (960133e) to head (f6b8345).
⚠️ Report is 451 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/utils/image_utils.py	80.00%	0 Missing and 1 partial ⚠️

❗ There is a different number of reports uploaded between BASE (960133e) and HEAD (f6b8345). Click for more details.

HEAD has 2 uploads less than BASE

Flag	BASE (960133e)	HEAD (f6b8345)
keras	5	4
keras-tensorflow	1	0

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #21683      +/-   ##
==========================================
- Coverage   82.63%   76.87%   -5.76%     
==========================================
  Files         577      577              
  Lines       59321    59325       +4     
  Branches     9300     9302       +2     
==========================================
- Hits        49020    45609    -3411     
- Misses       7913    11270    +3357     
- Partials     2388     2446      +58     

Flag	Coverage Δ
keras	76.74% <80.00%> (-5.72%)	⬇️
keras-jax	63.34% <80.00%> (+0.02%)	⬆️
keras-numpy	57.58% <80.00%> (+0.02%)	⬆️
keras-openvino	34.29% <0.00%> (-0.01%)	⬇️
keras-tensorflow	?
keras-torch	63.64% <80.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[fchollet]

Thanks for the PR! Please fix the code format and address the relevant comments from Gemini.

[fchollet]

You can fix the code format via sh shell/format.sh.

[hertschuh]

You're checking _format, but you're no longer verifying file_format.

[hertschuh]

It would be easier to do this line 186 instead of having to check both line 197 and lines 188-190.

in fact, you don't need _format at all, use file_format lines 181-185.

[fchollet]

This would be breaking

[utsab345]

Any update @fchollet ??

[fchollet]

This change would be breaking since it switches .jpg for .jpeg, changing the filenames the user expects on disk. We can't merge it as-is.

[hertschuh]

Thanks for the update.

[hertschuh]

Can you move this code to a keras/src/utils/image_utils_test.py file?

This should be fast enough, it doesn't need to be an integration test.

Then use the standard patterns:

• create a class that extends testing.TestCase
• use from absl.testing import parameterized instead of pytest.mark.parametrize
• use self.get_temp_dir() to get a temp dir

[hertschuh]

You can now remove this file.

[hertschuh]

You can now remove this file.

[hertschuh]

This is not covering the case that you added support for. That is, the case when the file_format was not provided and it was inferred from the path.

[hertschuh]

This can be simplified. There is no need for 3 variables for the file format. You're no longer using file_format after assigning to save_format, and you don't need save_format as it can be replaced by internal_format. I think the version you had earlier was correct.

[hertschuh]

Thank you for the fix!