I'm using OpenCV, for acquisition and presentation, in a portable program running in Windows or Linux.
I propose two functions, namely training and recognition, with a multithreaded architecture for recognition algorithms.
A video window would show the acquisition stream, with overlapped the results from either training or recognition, depending on the current function.
A minimal user interface - text console based - would allow to
- confirm training set, then run recognition
- return from recognition to training
Hope you're interested
Ing. Carlo Capelli