Ever AI vs. Industry Leaders
For our partners, a 1% error rate means billions of dollars in lost revenue.
Face recognition models are only as good as the data on which they are based. Unlike the competition, our models are only trained on proprietary datasets comprised of hundreds of confirmed labels per identity.
But don't take our word for it. Here are the numbers that prove it.
Note: Apple and Google don't offer face recognition services for enterprises.
* Accuracy: probability that two given faces are correctly identified as the same person
** Area Under Curve (AUC): probability that the model will rank a randomly chosen positive instance higher than a randomly chosen negative one
*** Numbers based on Labeled Faces in the Wild (LFW) image dataset
**** Competitor numbers based on publicly available APIs
***** IBM Watson metrics based on image comparison API (face verification not publicly available)
SPEED (API RESPONSE TIME)
Ever AI provides a self-hosted, on-premise API that eliminates additional networking latency and transfer time. Achieve even faster speeds with our industry-first SDKs.
* Speed: the time it takes the API to respond with a comparison score between two or more faces
** Clarifai: response time likely understated. API returns feature embeddings only and does not capture latency for distance calculations.