Show HN: How attractive is your website? Check using Visual Mind AI

Re:Show HN: How attractive is your website? Check using Visu... by bArray on Saturday September 19, 2020 @10:46AM (#24527759)

@myraahio Given how quickly your site runs, I'm not entirely sure it runs a neural network - seems more like some weighted analysis based on some defined measures on simple algorithms? I'll take a wild stab and guess that you're doing some analysis, some manual labeling and k-means clustering?

I would really consider running a light weight network over a small image (down-scaled to something like 320x200), if you're only binning classes then you should be able to get it to <1ms on a low-power CPU. Given how much web pages vary, you'll likely need upwards of 10k training images.

Consider having some feature on your site to allow users to help rate pages (with moderation oversight).

It's a good start anyway! I look forward to seeing this in the future!

Edit: Also consider some expansion opportunities - rating screenshots of software interfaces, mobile phone apps, etc, etc.

[ Parent | on HN ]

Re:Show HN: How attractive is your website? Check using Visu... by myraahio on Sunday September 20, 2020 @03:08AM (#24532409)

@bArray really appreciate your feedback. Especially expansion opportunity.

DL model is trained on the GPU - heavy AWS instance..end point rests on smaller instance. Sometimes it gives faster results when you run a URL which it has already processed. It will happen in case of famous sites. For new sites it need to process the info so it takes about 10 seconds to create report.

[ Parent | on HN ]