Photo.net Dataset - March'09 Release ------------------------------------ Ritendra Datta (datta [at] cse [dot] psu [dot] edu) --------------------------------------------------- Description of columns of photonet_dataset.txt ---------------------------------------------- Column 1: Index Column 2: Photo-ID Column 3: Number of aesthetics ratings received Column 4: Mean of ratings Column 5: Standard Deviation of ratings Columns 6-12: Distribution (counts) of aesthetics ratings in 1-7 scale, 6 being for 1, and 12 being for 7. NOTE: In a few cases, only the mean was available, the individual ratings were not available. In these cases, only the mean should be used. Columns 5 to 12 will be -1 in these cases. If your model involves distributions or std. dev. of ratings, please handle these cases as appropriate. How to obtain the original images? ---------------------------------- The photo ID is unique, but the exact URL of the files may change on the Photo.net server. Probably the best way to obtain the image URL is to download/wget the html page hosting the image, and scrape that html file for the image URL (there is usually a pattern that can be easily detected). The URLs for the html pages are currently of the following form: http://photo.net/photodb/photo?photo_id= e.g., http://photo.net/photodb/photo?photo_id=1008958 Instead of doing the above, at the time of this release, the images can be downloaded directly from URLs of the following form: http://gallery.photo.net/photo/-md.jpg e.g., http://gallery.photo.net/photo/1008958-md.jpg Copyright Issues ---------------- Please respect copyright that applies to you and to the data sources. Please do not re-distribute the image data. This is the same reason we do not distribute the original images.