Sorts coins by solenoid on a conveyor by classifying images with Caffe & DIGITS
To understand where this system is heading in the short term see the prototype CoinSorter's issues.
This group of programs and scripts is currently just a quick proof of concept to show physical coin sorting. It's currently sorting about 2 pennies a second, continuously. One solenoid and 2 physical bins are currently set up. Using Caffe it’s easy 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 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. This can be done using using DIGITS with default setting of AlexNet with no programming involved!
Click to watch a video of it in action:
View parts list and documentation for the conveyors and the CAD files.
For a long term project direction check out the issues for future milestones.
#CoinSorter Project:
- This program watches a directory for images to appear
- The image is sent to DIGITS (The Nivida Caffe front end) for classification
- Finally a solenoid turns on for coins that meet a certain criteria
#CanonSDKTutorial-noexe:
- Saves image captures to file and calls a MATLAB script to crop the coins to 256x256 images
- Originally published by Johannes Bildstein on CODE PROJECT
This project is project is brought to you by:
GemHunt.com: a for profit company helping people learn robotics, machine vision, and deep learning while providing DIY small part handling systems.
Thanks!
Paul Krush