Yes and no.
You will not be happy with having 1000 variables in your data set.
To implement this kind of survey, it is much (!) more elegant to use a multi-level structure.
And, of course, a random generator will support you with drawing 50 pictures per participant, while making sure that every picture is reated evently often. With randomUse() there is even an option to handle respondents (=count their ratings) who droped out after 30 pictures, or so.