Welcome to experimenters.

[BarryDaniel53]

This is very much a prototype for something never attempted elsewhere to my knowledge. Please ask questions or mention bugs and I will attempt to help out where possible. I do not profess to be a programming guru, as evidenced from the code. Good luck with your experiments. The application is very much limited by the hardware so some niceties have been impossible to implement. I welcome feedback on your testing so useful improvements to the code and documentation can be made. [email protected] or add a comment.

[penguin-sam]

This has great potential. I tried something similar years ago but there wasn't the tech and software to make it feasible. There are big problems such as changing light conditions; relative movement of camera, hidden areas, etc. However I wonder how much computer vision and ai stuff is now out there to be leveraged? I can see how cheap hardware really helps.

Recognising a small number of locos and wagons/coaches as they move over a fixed layout of tracks is a wonderful state of the art senario.

I have worked in AI and computer vision (albeit a little while ago ) I am an keen railway modeler and current software developer. I'm off to look into this!

PS What are the goals?

[BarryDaniel53]

You are correct, there are problems to address so the applications are limited to situations where the problems are manageable. Cheapness allows multiple cameras but it also brings poorer image quality to contend with. This software is not magic. It uses a “low” power processor and no AI techniques, just standard logic and is processor limited. Loco identification will require other methods. It is an experimental introductory level to video control, and cannot aspire to greater things at the $10 price point. However the reliability and interfacing to other control systems besides EX-CS can be further developed.