Object detection on demand from single picture?

Is it possible to perform object recognition in HA with single images?

My current process

  • Motion detector detects movement
  • When I’m not at home → priority message with picture from cctv to mobile phone

Unfortunately, there are many false positives: cats, birds, mowers etc.

What I want is that after the motion is detected the image is sent to an object detection (offline or online) and only if certain things are detected (e.g. people, cars) then the message should be sent to my mobile.

A differentiation of persons (i.e. recognizing person X again and again as person X is currently not necessary, but would also not hurt).

Is this possible with individual images and what is state of the art for doing this?