Hi @katiemn, reviewers, and future readers!

I'm Gemini, and I'm currently reviewing this pull request. In the meantime, I'm providing a summary for everyone to quickly grasp the changes and intent.

This pull request, titled "feat: imagen 3 inpainting and outpainting samples", aims to add new samples demonstrating Imagen 3's image editing capabilities, specifically inpainting and outpainting.

Here's a breakdown of the changes:

• New Python sample files added:
  • generative_ai/image_generation/imagen3_editing/edit_image_3_inpainting_insert_mask.py: This sample shows how to perform inpainting by inserting an object into a masked area of an image using a user-provided mask. The example inserts a "red hat" into an image of a woman. (Lines 1-80)
  • generative_ai/image_generation/imagen3_editing/edit_image_3_inpainting_insert_mask_mode.py: This sample demonstrates inpainting using automatic mask modes ("background", "foreground", or "semantic") to insert objects or backgrounds. The example adds a "beach" to the background of an image of a woman. (Lines 1-81)
  • generative_ai/image_generation/imagen3_editing/edit_image_3_inpainting_remove_mask.py: This sample shows how to remove an object from an image using a user-provided mask. The example removes a player from a volleyball game image. (Lines 1-81)
  • generative_ai/image_generation/imagen3_editing/edit_image_3_inpainting_remove_mask_mode.py: This sample demonstrates removing objects or backgrounds using automatic mask modes. The example removes the foreground (woman) from an image, leaving only the background. (Lines 1-80)
  • generative_ai/image_generation/imagen3_editing/edit_image_3_outpainting_mask.py: This sample shows how to perform outpainting, expanding the image content beyond its original boundaries using a user-provided mask. The example expands a picture of roller skaters to include a city skyline. (Lines 1-80)
• Corresponding test files added: Each of the above sample files has a corresponding test file (e.g., ..._test.py) to ensure the functionality works as expected. These tests verify that the image editing functions produce outputs with a byte size greater than 1000 bytes. (Lines 1-42 for each test file)

All new files include standard copyright headers (Lines 1-14 in each file).

I'll have a more detailed review shortly. In the meantime, please let me know if you have any questions.

And now, a little haiku to brighten your day:

Pixels shift and change,
Images reborn, anew,
Code's magic unfolds.

These samples demonstrate useful Imagen 3 editing features. The code is well-structured and easy to follow. The use of descriptive variable names and comments makes the code easy to understand. The test cases are also well-written and cover the main functionalities of the samples. A few minor suggestions for improvement are included below.