Teach AI dataset to identify specific objects on images

I need a solution where a well-trained AI can detect difference between images of:
– a person holds weapon in a crowded area
– no person holds weapon in a crowded area

It is something like a security check of a CCTV videostream.
You can choose any favourite platform, but it needs to be trainable further.

At first, I need to know an approximate time and price for this solution.