Video summary AI

I know this would be a huge task and very difficult, but it would be cool if you would create an AI that could look at the videos and create a summary text for each one, about what (if anything) was going on. It would be like anomaly detection, but with a description of the incident.

Example descriptions would be “Red car driving slowly”, “Person walking down the sidewalk”, “Garage door opening”, “Deliveryman leaving package”, etc. You could filter events by these labels too, or have it alert you to any incident involving certain labels, etc.

There would also be detection for anomalies like car accidents, fires, robberies, shootings, etc.

This would be extremely complex but it sounds cool!