Artificial Printing Defect Image Dataset (APDID)

Artificial Printing Defect Image Dataset (APDID) for image quality assessment

We created a dataset for printing quality assessment (PQA). This dataset contains  81 pure print images, each degraded by (**個) distortions in 5 levels (, which has totally (計**枚) images). Through a process known as crowdsourcing, we conducted a subjective PQA study on APDID, and then obtained (at least) 30 degradation category ratings (DCRs) per image.


Download: available as soon as possible (2022.8.5)
Paper: available as soon as possible



Pure print images81 print images, which are artificially generated by transforming the color profile from RGB to CMYK (Adobe ...). The original 81 images are obtained from Pixabay, which are the same original images as KADID-10k.
Defects1. Vertical black line
2. Vertical white line
3. Horizontal black line
4. Horizontal white line
5. ...
Scoring metric5-point DCR (Degradation Category Rating)
Subjects1200 cloud workers
# of subjects per image(at least) 30



Example of images (coming soon…)