Obviously there would be limits as to how much could be manually reviewed by hand (if 1000 people reviewed 1000 images each, only 0.02% of the images would be r

Obviously there would be limits as to how much could be manually reviewed by hand (if 1000 people reviewed 1000 images each, only 0.02% of the images would be reviewed assuming no overlap was required), but I wonder if there would be any benefit to attempting to crowdsource captions for the dataset for the worst available images