pkrush
Tenderfoot
- Joined
- Aug 26, 2015
- Messages
- 6
- Reaction score
- 3
- Golden Thread
- 0
- Location
- Carol Stream, IL, USA
- Detector(s) used
- Caffe Deep Learning Framework
- Primary Interest:
- Other
Hi,
I upgraded my Ryedale to sort out the wheat pennies. Well sort of. I choose not to use the coin discriminator's solenoid, but I could have. This system is currently sorting about 2 pennies a second, continuously. One solenoid and 2 physical bins are set up. Yes it's missing wheats that come out heads up. I am planning to deal with this, but for now I am just happy it works as well as it does.
Why did I build it? I would like to design a low cost and easy to build DIY coin sorter. To help with this design process, I built this proof of concept you see in the video below. Let me know what you think of it...
Be sure to check out the video at the end of this post.
CAD files, Code, and future plans are at:
https://github.com/GemHunt/CoinSorter
For those of you who are technically inclined:
It uses a Caffe deep learning framework out on AWS GPU instance to distinguish between designs of coins. For example you can train a convolutional neural network (CNN, what Caffe uses) to determine if a coin image is heads vs tails or say to recognize the state on a random US state quarter image. Using the "copper" image set out on GemHunt.com, Caffe can tell heads vs tails between US copper pennies 99.9% of the time. I have been doing this using NVidia’s DIGITS web interface to Caffe with the default setting of AlexNet.


I upgraded my Ryedale to sort out the wheat pennies. Well sort of. I choose not to use the coin discriminator's solenoid, but I could have. This system is currently sorting about 2 pennies a second, continuously. One solenoid and 2 physical bins are set up. Yes it's missing wheats that come out heads up. I am planning to deal with this, but for now I am just happy it works as well as it does.
Why did I build it? I would like to design a low cost and easy to build DIY coin sorter. To help with this design process, I built this proof of concept you see in the video below. Let me know what you think of it...
Be sure to check out the video at the end of this post.
CAD files, Code, and future plans are at:
https://github.com/GemHunt/CoinSorter
For those of you who are technically inclined:
It uses a Caffe deep learning framework out on AWS GPU instance to distinguish between designs of coins. For example you can train a convolutional neural network (CNN, what Caffe uses) to determine if a coin image is heads vs tails or say to recognize the state on a random US state quarter image. Using the "copper" image set out on GemHunt.com, Caffe can tell heads vs tails between US copper pennies 99.9% of the time. I have been doing this using NVidia’s DIGITS web interface to Caffe with the default setting of AlexNet.


Amazon Forum Fav 👍
Upvote
0