How to Conduct Simple Comparison with Subjectify.us
Dataset
In this video we show how to conduct subjective comparisons with Subjectify.us. For this illustration, we perform a basic comparison of image-upscaling methods using the Subjectify.us platform. Our comparison involves three well-known methods:
- nearest-neighbor
- bilinear
- bicubic.
We apply them to two test images at three upscale factors: 2 times, 3 times and 4 times.
 |
 |
|---|---|
To generate the test dataset we prepared simple Python script.
Perceptual Data Collection
48 participants took part in the study. Each study participant compared 33 image pairs including 3 hidden quality-control comparisons between ground truth and a nearest-neighbor method; answers of 1 participant were rejected, since this participant failed at least one quality-control question.