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 2nd Proof of Concept issues and the Prototype Coinsorter Issues
Check out the starting point issues to see what is currently being worked on.
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: helping people learn robotics, machine vision, and deep learning while providing DIY small part handling systems.
Thanks!
Paul Krush