I moved last year and now have the chance to construct my long-dreamed-of basement empire. There are TV’s scattered around the basement, the idea being to show things like slide shows, railroad DVD’s, Virtual Railfan and live video from a CC video surveillance system I have. This got me to thinking about feeding that live video stream to my computer which could send it through a computer vision processor like YOLO. I was wondering if anyone else had experimented with using AI and live video basically as block occupancy detection? The simple start is to do something like. “Hey, the computer sees a train headed for the grade crossing at Elm St. Activate the crossing gates at Elm.” Then we move to more complicated things like, “Hey, that train stopped short of the crossing, put the gates back up.” Then really complicated things like, “Hey, the ditch lights came back on, that train at Elm St is about to move again. Drop the crossing gates until it passes through.” Eventually it would be nice to get to the point of being able to do crazy things like, “Hey, there are three people in the aisle and they are all looking left. Sound a whistle behind them so they all look right. Then start some animation sequence that will amaze them.” Has anyone experimented with using vision technology on camera feeds from their layout?
3 Likes
What you describe sounds amazing, but for someone of my age and limited intelligence, I’m afraid I couldn’t even begin thinking of a system like that. If you ever get even a small amount of a system like this working, please post photos and videos on the forum for us to see!
5 Likes
Well, I don’t really think too much of AI but it could maybe have some sort of an application here. I’ll stick with using automatic third-rail block detection systems for now.