Face/eye detection in thumbnails

[deanmoses]

Can we automatically detect where all the faces/eyes in the image are and crop the thumbnail such that it includes all of them?

It wouldn't zoom in - it would crop the biggest square possible. So on landscape photos it would just change the x axis of the crop.

Would this happen in the uploaded image processor or on-the-fly when requesting a thumbnail that doesn't have crop info?

If it happens in the uploaded image processor, it would save the crop info to DynamoDB just like a manually selected crop, and also save that it was cropped using face detection, so that if I re-apply a better algorithm later it knows to update it.

There's an AWS Solution template that does this with Sharp + AWS Rekognition: https://aws.amazon.com/solutions/implementations/serverless-image-handler/