I'm always fascinated by the use of incentives to aggregate large, useful datasets. This one by @codatta_io was v cool: 1.2 million real, validated food photos with ingredients & nutrition labels Here's how they did it in 6 weeks: