Show HN: Analyzed 30k images from 300 art blogs to find the most popular art

arthunted.com

17 points·chrischen··3 comments

You describe your site as "Every day we analyze over 30,000 images from over 300 blogs to surface trending artwork, photography, and design.", but you've gone some way to show that there's no such thing as "trending artwork". Analysing 300 blogs and finding that the most popular images are only featured on 9 of them demonstrates that there isn't a trend that sites follow, or if there is it's indistinguishable from randomness.

It crawls 300 sites and scans the images from just the front page, and it does this daily. 30k images includes a bit of fluff, which is trimmed down, so the actual real images from articles is a bit smaller. This was also from the initial scan, so a daily yield of only the newly discovered images is probably only around 5k.

The crossover would be a lot larger if I ran this from say a week of submitted articles (whereas right now it's a crossover of basically front page articles from 1-2 days).

If you compare this to hypem.com, the collisions are actually as frequent if not more, considering hypem.com gets updated about once a week.

Brief summary: Did a shallow scan of just the front pages of 300 art blogs. Initially pulled in about 30k images, with a daily additions being about 5000.

I used pHash to dedupe images and tally votes for every unique "art" piece.

Reply on news.ycombinator.com