Amazon Web Services announced a few really interesting services during re:Invent last week. One that caught my attention was AWS Rekognition.
What does Rekognition do? Well, it provides developers with the ability to understand images with confidence. But, don’t take my word for it, here is a simple test application I published today.
I might have published sooner, but I was busy studying for the AWS Architect Associate Level exam (which I passed just this morning).
What will we use Rekognition for at Link Digital? Well, we have a few relevant projects already in the queue for the semantic markup of image assets. Based on our extension of CKAN as a digital asset management system, which we’ve demonstrated at data.wearecbr.com.au (Drupal experience) and data.wearecbr.com.au/data (CKAN experience), we’re thinking that the extension published as CKAN Galleries is now primed and ready for integration with our benevolent and time saving AI overlords.
But really, how helpful is rekognition? Here is one example taken from the photo library of our Exectutive Director, Steven De Costa, who was in attendence at re:Invent:
The picture was taken during the re:Invent partner day, during the end of the morning session where the CEO of Amazon Web Services, Andy Jassy, was giving the 4,000 odd attendees schooling on what customers need (hint: partners with more trained and certified staff, like me). The value which rekognition provides is a shortcut on the markup of such an image, suggesting perhaps the most obvious tags which could be associated with it with confidence, along with a longer tail of lower confidence tags (for this image there were another 15 suggestions with confidence lower than 90%).
However, the fun is in playing with rekognition yourself, which is why I put together the simple test script you can now use. If you’d like to contribute and improve this basic test page, please check out our rekognition repository on GitHub.com. And remember, I’m traditionally a SydAdmin rather than a coder, so I’m very appreciative of any suggestions delivered via issues or pull requests.
Two months later I got my third and final associate level certificate, the SysOps Administrator associate.