Basically when you setup the machine and train it to recognize/compare against sample models vs the camera scanned brass headstamps when you are sorting. From Seth - It uses Google TensorFlow model "Inception". If the deep learning option is selected it uses ML.net from Microsoft. Once a positive ID is completed and rated against the confidence level you specified in the application, it drops the headstamp identified in the bin you specified. Anything it cannot identify can just drop in the catch all bin, you can re-run these again or use it to further train the AI. Saves time manually sorting. :)
Finally, a use for AI i can get behind. But is that not closer to a Machine Learning task?
Basically when you setup the machine and train it to recognize/compare against sample models vs the camera scanned brass headstamps when you are sorting. From Seth - It uses Google TensorFlow model "Inception". If the deep learning option is selected it uses ML.net from Microsoft.
Once a positive ID is completed and rated against the confidence level you specified in the application, it drops the headstamp identified in the bin you specified. Anything it cannot identify can just drop in the catch all bin, you can re-run these again or use it to further train the AI. Saves time manually sorting. :)