This supplemental web site provides a summary of our automatic macro results for 20 manipulations: 7 face, 8 landscape, and 5 global.
You can click on the results link for each manipulation to see a sample of our results (both successful and less successful).
Comparisons with Grabler09: We compared our results against the technique of Grabler09 for the manipulations that only contain selection and parameter adjustment operations (Contrast, Skin Tone, Black and White, Cross Processing, Lomo, Typographic Effect). We did not include comparisons with Grabler09 for manipulations with brush strokes, since Grabler09 cannot transfer strokes. Grabler09 also does not take into account the parameter settings for multiple demonstrations. Rather than just copying parameter settings from the first demonstration, we modified the Grabler09 technique to use the average parameter values across all training demonstrations.
Comparisons with Average: We also compared our results against the Average technique (described in the paper) for all manipulations that we evaluated in our Mechanical Turk study (Bag Removal, Contrast, Dark Sky, Eye Makeup, Film Noir, Lomo, Sunset Enhancement). In addition, we included Average results for all other manipulations that do not contain any brush strokes (Skin Tone, Black and White, Cross Processing, Typographic Effect) because we wanted to specifically evaluate whether our technique for learning adjustment parameters produces better results than just averaging parameter values.
Note that this website is best viewed in Firefox.