Vufind Recognize is an algorithmic computer vision/object recognition API, i.e no human crowd-sourcing layer is involved. There are pros and cons to that. Major pros are: It's real-time, and scales very cost-effectively including for video. However, at the same time, it can't be expected to be 100% accurate. Please note that it's well documented that humans can't deliver 100% accuracy either if they have to inspect 20 images or more per second, or if you expect them to tag all the objects in the frame, and in case of video they tire very quickly. Most importantly, crowd-sourcing costs orders of magnitude more than machine vision.
Vufind's platform is quite configurable and shortly we will expose parameters to enable you to choose to optimize for maximum "recall" (show as many tags as possible even if not very confident), or minimize false-positive rate (i.e only show a tag when the engine is very confident). Our recommendation is to choose minimize FP rate for consumer apps, such photo-albums and augmented reality apps. On the other hand, for advertising/commerce applications, maximizing recall might be a better choice.
We tend to refer to them as "object classifiers", although researchers distinguish between category recognition for a class of objects with variety of shapes (for ex. car, high-rise building, long-dress, sashimi, cat) vs. instance recognition (golden-gate bridge, EmpireState Building, salmon sashimi)
Feb 15, 2014 Update:
Version V1.7.5 of public developers version API Recognize endpoint has been updated, and now has the following classifiers below. We've always adding new classifiers to the set based on customer demand, so feel free to email us any time with the needs of your application. We'd love to hear from you. firstname.lastname@example.org
Genre =fashion: -Long-dress -Short-dress -Purse -High-heels -Women-boots -Women-Long-coat -Sunglasses
Genre=decor - modern-decor - living-room - dining-room - bedroom
Genre= brands: - Pepsi - TacoBell - coming soon (LouisVuitton)
Genre= scenes - outdoors - beach - creek - garden - highrise-bldg - mountain - playground - sea/ocean - shoppingMall
Genre=food - pizza - fried-rice - fried-dumplings - salad - sushi - sashimi - shrimp-cocktail
Genre=landmarks -eiffel-tower -empirestate-bldg -goldengate-br -bay-br -disneycastle -oslo-cityhall -sonoma-cityhall -sydney-opera -white-house
Genre=man-made - sedan - cruiseship - sportscar
- No information
Simple & Straightforward Pricing
Pay as you go. No long-term contracts.
$0additional fees may apply
1,000 / mo.
$0.0500 per extra
50,000 / mo.
$0.0150 per extra
300,000 / mo.
$0.0120 per extra
1,500,000 / mo.
$0.0100 per extra
specifies the types of objects you are interested in getting tags for. For example, for location context, you'd specify genre=landmarks, for shopping apps, use genre=apparel, etc. Supported genres: fashion, decor,food,landmarks,man-made,moms,nature,scenes.
The URL of the photo, preferably terminating in .jpg or .png. We do not accept .GIF images
your app's id of user associated with this image (for interest graph purposes)
Vufind's API key
Vufind's persistent token
string – Server overloaded Please try again a bit later. Thanks!HTTP 502
string – Server OverloadedHTTP 503
string – Bad API parameter values, and possiblky the image/video URLHTTP 400