[P] using machine Learning to build robot to pick up nails at the river : MachineLearning

Hello,

I have a dream. I run a local cleanup group that focuses on collecting nails from burnt pallets and other waste at our river. This is a major problem. So far I have picked up 6,500lbs of nails. I have built 2 magnetic sweepers. They are custom built with the strongest magnets possible. So here is the dream: When the nails hit the stainless steel tray ((the bottom)(the magnets sit in side the tray and swings out to drop the nails from the bottom) they make a very distinctive sound. Once the tray has been saturated with nails the sounds dissipate. It is then time to dump the nails in collection pile. This is very labor intensive volunteer work. It is easy to identify hot spots of nails due to the sound it makes clanging against the tray. Is this type of audio input good enough to train an AI to cover a large area and identify nail hot spots? The robot part seems easy enough… its the training the AI that has me lost. Is there anyone who might have ideas on how I can do this or help with the project? Once it works on one then we can scale it up. Illegal dumping and pallet burning is a nation wide issue that caused countless flat tires, injuries to wildlife and pets.