MICE User Study

Welcome! Your help is vital in evaluating multi-instance image editing models. Please follow these instructions:

  • Source Information: On the left is the original image. On the right is the Target Guide showing the bounding boxes to edit with the corresponding text target or object target.
  • Model Comparison: You will see two generated edits from different models.
  • Your Task: Choose the image that best followed the instructions while maintaining high visual quality. Consider the preservation of the writing style or object style from the original image and the capabilities of the model in attempting to edit all the target instances.
  • Controls: Use your Mouse or Arrow Keys (Left/Right) to vote. We estimate 15 minute time duration with 8 seconds per sample.

Thank You!

You have successfully completed all the samples in this study.

Total Votes: 0

Your contributions are greatly appreciated and will help us improve multi-instance image editing models.

User: Anonymous
Progress: 0% (0/0)

Comparisons

Source Information (Left: Original image with bounding boxes, Right: Bounding boxes with target text/object (Adapted for your visualization)
Source Image
Which model followed the instructions better? Look for: Style preservation from original image, Nr. of attempted edits, and correct edit)
Option A
Left Arrow
Option B
Right Arrow
×

Top Contributors