buckyb

I'm looking to get pointed in the right direction. I have a couple of IP security cameras (Reolink RLC-423) set up with general, area-based motion detection at home. When motion is detected the files are saved to a server using Synology's Surveillance Station. Everything works great but many of the files are triggered from what I would consider uninteresting events (e.g. bugs flying across the field of view) that I can't eliminate by reducing the motion detection zone or sensitivities. It's pretty straightforward (but very time consuming) to review the files and determine if the video contains something I'm interested in (like a car or person or animal). I was wondering if there's a machine learning package that I could point to a directory of video files and have it determine which files have something of interest to me. I could produce a training dataset containing hundreds of video files that I have predetermined are "interesting". I'm not looking for anything real-time, so hopefully that will help. I know enough Python and Linux to get into trouble, so if some compilation/customization of the software is required that should be ok. Thanks for any help.