We've been playing with Amazon Rekognition service which can analyse video or still images for faces and then check a database of millions of images in a couple of seconds for a match.
With Amazon Rekognition, you can identify thousands of objects (e.g. bike, telephone, building) and scenes (e.g. parking lot, beach, city). When analyzing video, you can also identify specific activities happening in the frame, such as "delivering a package" or "playing soccer".
When analysing faces from a still camera we were able to get a 80% hit rate, the camera needed to be straight on and not profile.
This technology took just a few minutes to deploy on AWS, it gives you a web interface to upload the images into a database and another to upload the video of image that you want to look for. - simple to deploy and easy to use!
It could easily be used by Sports Stadiums to reduce trespassing, but picking people out in a crowd has a little way to go before it's more acurate and useful.