Tweet
Hi, I'm new to the forum and this is my first post, so if this topic has come up before please point me in the right direction. I am interested in building a neural network based (hardware, not software based) metal detector. Neural networks are great at pattern recognition and (once trained) perform well even with missing or corrupt data. This machine would have a "train" and an "operate" mode, so if you get an interesting signal you can capture a training cycle (probably 2x E/W and 2x N/S sweeps) or just use it in the operate mode.
I conceive of building a "training set" with many thousands of 1-second sweep elements (with collaboration it could easily be over a million samples and with file sharing it would continue to grow each year!) I am an engineer capable of designing the circuits, PCB, packaging, face-plate layout, etc. It might incorporate an Arduino and a NeuroShield - not sure, but that might be a place to start. We can always refine things and make changes as new hardware become available.
Any experience or ideas regarding a project like this?
Rick
I conceive of building a "training set" with many thousands of 1-second sweep elements (with collaboration it could easily be over a million samples and with file sharing it would continue to grow each year!) I am an engineer capable of designing the circuits, PCB, packaging, face-plate layout, etc. It might incorporate an Arduino and a NeuroShield - not sure, but that might be a place to start. We can always refine things and make changes as new hardware become available.
Any experience or ideas regarding a project like this?
Rick
Comment