Add dtype tests for ops.image.*.

[james77777778]

This PR adds dtype tests for ops.image.* and includes fixes discovered.

[gemini-code-assist]

Summary of Changes

Hello @james77777778, 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 focuses on improving the robustness of image processing operations by adding extensive data type (dtype) tests and implementing necessary fixes. The changes ensure that image manipulation functions maintain correct data types throughout their execution, especially when dealing with various floating-point precisions. This enhancement is crucial for maintaining numerical stability and consistency across different backend frameworks.

Highlights

• New Dtype Tests for Image Operations: This PR introduces comprehensive dtype tests for various image operations within keras.src.ops.image, ensuring that these functions correctly handle and preserve data types across different backend implementations (JAX, NumPy, TensorFlow, Torch).
• Dtype Compatibility Fixes Across Backends: Several fixes were implemented across backends to address dtype compatibility issues, particularly for half-precision floating-point types (float16, bfloat16), which often require casting to float32 for underlying library calls (e.g., scipy.ndimage, np.linalg.solve, torch.linalg.solve) and then casting the result back to the original dtype.
• JAX Backend Dtype Adjustments: Specific adjustments were made in keras/src/backend/jax/image.py to ensure correct dtype propagation in gaussian_blur and kernel creation.
• NumPy Backend Half-Precision Support: Significant changes were applied to keras/src/backend/numpy/image.py for affine_transform, compute_homography_matrix, map_coordinates, and gaussian_blur to robustly handle float16 and bfloat16 by temporarily upcasting to float32 for SciPy/NumPy functions that lack direct support for these types.
• TensorFlow and Torch Backend Dtype Consistency: The keras/src/backend/tensorflow/image.py and keras/src/backend/torch/image.py files received updates to ensure proper dtype handling for input tensors and intermediate calculations in functions like gaussian_blur, perspective_transform, and compute_homography_matrix.

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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

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 adds comprehensive dtype tests for the ops.image module, which is a great addition for ensuring correctness across different backends and data types. The associated fixes in the JAX, TensorFlow, and PyTorch backends correctly address dtype consistency issues. However, I found a bug in the NumPy backend's compute_homography_matrix function where a dtype casting operation has no effect, leading to an incorrect output dtype. I've provided a suggestion to fix this.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.48%. Comparing base (b9ff57a) to head (5bca106).

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #21612      +/-   ##
==========================================
+ Coverage   82.45%   82.48%   +0.03%     
==========================================
  Files         572      572              
  Lines       57337    57358      +21     
  Branches     8970     8969       -1     
==========================================
+ Hits        47277    47314      +37     
+ Misses       7761     7753       -8     
+ Partials     2299     2291       -8     

Flag	Coverage Δ
keras	82.29% <100.00%> (+0.03%)	⬆️
keras-jax	63.57% <14.28%> (-0.01%)	⬇️
keras-numpy	57.89% <64.28%> (+0.02%)	⬆️
keras-openvino	34.33% <0.00%> (-0.02%)	⬇️
keras-tensorflow	64.22% <7.14%> (+<0.01%)	⬆️
keras-torch	63.80% <35.71%> (+0.01%)	⬆️

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.

[shashaka]

Great work on expanding dtype coverage and strengthening consistency across ops.image.*.

[shashaka]

@fchollet Since I don’t have merge rights, could you please check this PR and merge it?