AWS Rekognition Performance

Question

I'm having trouble finding any detailed performance specs on AWS Rekognition. Would anybody be able to point me in that direction? I've seen some metrics thrown around such as 1.5 second search time when using the SearchFaces API, but it doesn't state how many photos that search was against.

I'm looking to design something fairly large that will have approximately 100 million indexed photos in a collection. That collection will be compared against by approximately 25,000 new photos every day. Basically I'm just looking to determine how this algorithm compares against some non-cloud algorithms (ie. Cognitec, Neurotech, etc.)

Thanks!

Accepted Answer

Unfortunately, there's no simple answer to that question, as it depends upon complexity of the picture, among other things (e.g. > [blog post ](https://aws.amazon.com/blogs/compute/analyzing-performance-for-amazon-rekognition-apps-written-on-aws-lambda-using-aws-x-ray/#:~:text=Specifically%2C%20you%20can%20look%20at,for%20the%20random%20landscape%20image.)< that talks about more or less than 5 faces in a picture). Some tips for optimizing performance can be found in the dev guide > [here.](https://docs.aws.amazon.com/rekognition/latest/dg/operation-latency.html) <

AWS Rekognition Performance

関連するコンテンツ