Ask HN: Does anyone know what image recognition algorith the app “Foap” uses?

28 points·kebinappies··5 comments

Foap is an app that allows you to sell photos, but the system it uses for tagging is insane. It's crazy accurate and I want to know who made it.

View on HN

They use Clarifai

Source: https://www.clarifai.com/

> In order to connect brands with the images they’re looking for, Foap uses Clarifai to consistently tag all these photos to make them easily searchable.

Pricing: https://www.clarifai.com/pricing

More on Foap <-> Clarifai: http://blog.clarifai.com/uncover-buried-stock-photography-tr...

Google's Vision API FWIW unsurprisingly blows Clarifai out of the water.

https://cloud.google.com/vision/

I've checked Clarifai, Google, and Microsoft cognitive services with this photo https://farm4.staticflickr.com/3232/2492569957_1e5ebdec11_b_... (mariachis in a boat in Xoximilco lake) - this should be hard to classify, of course.

This are the results: NONE of them identified people as mariachis ;-)

Microsoft: Captions: "a group of people on a boat" Tags: outdoor, sky, tree, boat

Clarifai (TAGS): people, vehicle, water, festival watercraft, transportation system, tourist, travel tourism, city, adult, group boat, recreation, man, outdoors competition, canal, sfw transportation, travel

Google (tags): Boat 79% Vehicle 77% Boating 67% Waterway 58% Raft 53%

My only conclusion is that both Google, Clarifai, and Microsoft are doing a great job. You should take a look at what you need and choose the best provider.

Exif data may have location coordinates as well

FTR you can just build your own very easily out of any of the open source DL frameworks out there and their derivatives. E.g. digits and deepdetect will give you an API along with the pre-trained models.

Reply on news.ycombinator.com