Machine learning has come a long way, and one aree that has improved tremendously is image analysis. It is now possible to identify people and objects in photos at a scale that was previously impossible, and it makes something I have always wanted to do with TroopTrack a reality:
Searching for pictures of my kids. On a bridge. Or when they are smiling. Or when they are in the same room with a duck. Or in the Alps.
You get the idea.
Today I am pleased to announce a new BETA feature: TroopTrack Faces. TroopTrack Faces allows you to search pictures in your TroopTrack photo albums like never before.
Please keep in mind this is a BETA release, which means:
- It might crash
- It might be slow
- It might be unattractive
- It might not feel entirely complete
We want your feedback on this new feature. It is an opt-in feature, and for now you need to request it by sending an email to @KelsieC (firstname.lastname@example.org). NOTE: SOME PEOPLE ARE CREEPED OUT BY THIS KIND OF TECHNOLOGY. PLEASE MAKE SURE YOU TALK TO YOUR COMMITTEE BEFORE TURNING IT ON.
Okay, screen shots…
When you upload a photograph to TroopTrack we use an AWS service called Rekognition to identify faces and labels in the photo. If there is only one face in the photograph then it can be used to search for other photos of the person in the photo. This screenshot shows these one-person photos - you can click on any one of them to see more photos that include the person shown.
You can also filter by labels. AWS Rekognition will identify objects in the photo and add labels to it. For example, if it detects people wearing glasses, it will add the label “eyewear” to the photo. You can choose to filter by label in the top right of the Faces menu. Here’s me filtering by “bell tower”.
This technology isn’t perfect. Sometimes the mistakes are pretty amusing.
In the future I would like to add some other features, but I want to see how these two features do first. Here’s what I’m thinking of adding:
- Photo tagging
- Multi-label search
- Image moderation (blocks the display of inappropriate images)
We hope you like this new feature. Please let us know what you think.