These environments are the manually created input to our algorithm. They reflect the styles and object distributions the artist wants in the synthesized results.
Here we show the percentage of total hits for each possible rating for images in this category, and the cumulative percentages that are at least as good as each rating. This gives an indication of how many results you would have to search through before finding one of the desired quality.
5 | 4 | 3 | 2 | 1 | |
Percentage with rating | 38.33% | 17.5% | 34.16% | 9.16% | 0.83% |
Cumulative percentage with at least rating | 38.33% | 55.83% | 90.00% | 99.16% | 100.00% |
Below we show all of the images in this category that were presented to the participants in the perceptual evaluation study. For each image, we show the mean, standard deviation, and all the individual results.