Human Annotation Analysis

[292725651]

The annotations with prompt, classification, overall_rating, image_text_alignment_rating, fidelity_rating are released. Will the annotations of image issues (i.e. repeated generation, body problems, fuzzy images, toxic, pornographic, violent, or violating protected groups in your papers) be released? Thank you!

[xujz18]

Thanks for your attention! Please refer to #75 to find other annotation results.

[292725651]

Thanks for your reply. However, I found that 'problem_data.json' contains only prompts. Does it contains the detail classifaction of problem (e.g. 'xxxxxx.webp' has body problems, 'xxxxxx.webp' has pornographic problems)?

[crisnvg]

+1, I also couldn't find overall_rating, image_text_alignment_rating, fidelity_rating

[ctxya1207]

谢谢你的回复。但是我发现'problem_data.json'只包含提示。它是否包含问题的详细分类（例如'xxxxxx.webp'有身体问题，'xxxxxx.webp'有色情问题）？Did you find the file with the image problem annotations?

[xujz18]

The problem_data.json file, where you see the prompt and image paths belong to a dictionary with the dictionary key names conflict, limbs, unclear, toxic, sex, and violence, which are the desired problem categories.