Skip to content

Instantly share code, notes, and snippets.

@taesiri
Created February 5, 2025 20:38
Show Gist options
  • Save taesiri/32b128eefdd346c85bc4c96afe673823 to your computer and use it in GitHub Desktop.
Save taesiri/32b128eefdd346c85bc4c96afe673823 to your computer and use it in GitHub Desktop.

Feedback and Proposed Adjustments for Image Editing Categories

1. Re-coloring

  • Current Status: Definition is acceptable.
  • Feedback:
    • Add a note clarifying that users are assumed to use the “accurate definition” of colorize (i.e., converting black & white photos to color).

2. Relight

  • Current Definition:
    "Improve or change the lighting conditions of the scene such as the temperature, color, direction, or position of the light source."
  • Proposed Adjustments:
    Split into two distinct categories:
    • Relight – Temperature/Color:
      • Definition: "Improve or change the lighting conditions of the scene, focusing on the temperature or color."
      • Example: "Please get rid of the green light or change it to another color (like pink)."
    • Relight – Direction/Position:
      • Definition: "Improve or change the lighting by altering the direction or position of the light source."
      • Examples:
        • "Can someone relight this photo, completely removing all harsh shadows on the subject (face and body)?"
        • "I love this picture of me and my son but the light was coming from the wrong direction. Can someone make it better?"
        • "Can someone make lighting better / remove shadows?"
        • "Can someone please adjust the lighting so it isn’t so dark inside, but still bright outside?"
        • "Looking to make this person’s face less flashy—soften the harsh light from the spotlight on the stage."

3. Super-resolution

  • Current Status:
    Acceptable—assuming users correctly understand “resolution.”
  • Feedback:
    • Remove overly generic requests such as:
      "Can anyone help improve this picture?"
      (This request is too general and might be better served by an auto-adjustment rather than a resolution-specific operation.)

4. Adjust

  • Current Definition:
    Intended for overall adjustments to the image’s appearance.
  • Feedback:
    • Clarify that the definition applies to global adjustments rather than selective ones.
    • Note that examples like "Increase saturation a bit on the elephants" imply a selective adjustment, which may need a different treatment.
    • Ensure that requests mentioning “contrast” are directed to “Adjust” even if they also refer to lighting.

5. Delete

  • Current Status:
    Definition is clear.
  • Feedback:
    • Remove generic examples such as:
      "Can someone help clean up this picture of a picture?"
      (This example is too vague to be helpful.)

6. Crop

  • Current Definition:
    "Trim the edges of an image to make a smaller image to meet specific size requirements."
  • Feedback:
    • Clarify whether users must provide specific new dimensions.
    • Propose splitting into two definitions:
      • Crop (Standard Size/Ratio):
        • Definition: "Trim the edges of an image to achieve a specific or standard image ratio."
        • Example: "Crop to square format for social media."
      • Crop (Content-based):
        • Definition: "Trim the edges of an image by referencing the objects within the image."
        • Example: "Crop the photo to eliminate the space to the left and right of the elephants."

7. Add

  • Current Status:
    Definition is generally acceptable.
  • Feedback:
    • Confirm that any request containing the word “add” (even if it mentions lighting) is classified under “Add.”
    • Reconsider examples that might overlap with “Relight” (e.g., "Could anyone add realistic lightning effects to this photo?") and ensure they follow the rule that the term “add” takes precedence.

8. Replace

  • Current Definition:
    "Substitute objects or text in the image with something else while keeping the rest of the image unchanged."
  • Proposed Adjustments:
    Split into two groups:
    • Replace with an External Element:
      • Definition: "Substitute an object or text with something not originally in the image."
      • Example: "Please change the pamphlet she is holding into a dictionary."
    • Replace with an Internal Element:
      • Definition: "Substitute an object with another element from within the same image."
      • Example: "Please perform a face swap using the man in the yellow shirt and the man in the blue/black polo."

9. Apply

  • Current Definition:
    Indicates modifications applied to the overall appearance.
  • Feedback:
    • Some examples (e.g., "Add a Gaussian blur to the background") seem to involve selective adjustments.
    • Consider whether such selective modifications should be:
      • Re-categorized separately, or
      • The overall definition broadened to explicitly allow selective changes.

10. Zoom

  • Current Status:
    Both the definition and examples are acceptable.
  • Feedback:
    No changes required.

11. Transform

  • Current Definition:
    Involves changing the geometric properties of the image.
  • Feedback:
    • Clarify how “Transform” differs from “Crop.”
    • Distinguish cases such as:
      "Can anyone help me format this for Instagram—it won’t fit."
      (This example may require reformatting rather than simple trimming.)

12. Move

  • Current Status:
    Definition is clear.
  • Feedback:
    • Works as long as the user can clearly indicate the target location for the object.

13. Clone

  • Current Status:
    Definition and examples are acceptable.
  • Feedback:
    No changes required.

14. Undo

  • Current Definition:
    Involves removing elements, but may overlap with Delete.
  • Feedback:
    • Clarify the distinction between “Undo” and “Delete.”
      • For example, "Undo" might be intended to reverse a recent or accidental action, while "Delete" is a deliberate removal.
    • Examples like:
      "Can someone please undo my stupid doodles on this pic of my daughter?"
      may need re-evaluation to ensure they fit the intended operation.

15. Merge

  • Current Definition:
    "Combine multiple elements or effects from multiple images into a cohesive final image."
  • Feedback:
    • Differentiate “Merge” from “Add”:
      • Add: Inserting new elements that were not originally present.
      • Merge: Combining elements from two or more images.
    • Confirm that if a user provides two images (or requests combining elements from them), the operation should be classified as “Merge.”
      • Example: "Could someone merge these two pictures? I want JUST my mother (second picture, white dress) to be placed with my grandparents (first picture)."

16. Specialized Operations

  • Current Status:
    Definition and examples are acceptable.
  • Feedback:
    No changes required.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment