Hybrid cloud/camera AI computer vision system

A system comprising:

• a smart camera configured to (i) capture video data comprising a plurality of video frames, (ii) generate metadata in response to detecting a triggering event and (iii) select a plurality of images from said plurality of video frames of said video data in response to said triggering event and (iv) communicate said plurality of images and said metadata;
• and
• a server computer configured to (i) receive said plurality of images and said metadata, (ii) analyze said plurality of images to perform an additional detection corresponding to said triggering event, (iii) compare a result of said analysis of said plurality of images to said metadata to determine updated local parameters for said smart camera in response to said additional detection determining a false positive detection, and (iv) generate a signal in response to said additional detection, wherein
• (a) said server computer implements a first neural network for analyzing said plurality of images,
• (b) said smart camera implements a second neural network for detecting said triggering event in response to said local parameters,
• (c) said first neural network uses more computing resources than said second neural network uses for detecting said triggering event,
• (d) said signal comprises said updated local parameters for said smart camera in response to said false positive detection,
• (e) said updated local parameters are generated to update said second neural network to prevent said false positive detection, and
• (f) communicating said plurality of images uses less bandwidth than communicating said video data.