https://github.com/primaryobjects/digitrecognizer
Machine Learning digit recognizer (MNIST data-set) in C# .NET
https://github.com/primaryobjects/digitrecognizer
Last synced: 10 months ago
JSON representation
Machine Learning digit recognizer (MNIST data-set) in C# .NET
- Host: GitHub
- URL: https://github.com/primaryobjects/digitrecognizer
- Owner: primaryobjects
- Created: 2014-01-07T17:55:11.000Z (about 12 years ago)
- Default Branch: master
- Last Pushed: 2014-01-13T04:15:30.000Z (about 12 years ago)
- Last Synced: 2025-03-21T06:41:36.982Z (10 months ago)
- Language: C#
- Size: 19.1 MB
- Stars: 6
- Watchers: 2
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: readme.md
Awesome Lists containing this project
README
Machine Learning: Digit Recognizer (MNIST data-set)
--------
Read the full article at http://primaryobjects.com/CMS/Article159.aspx
This is a machine learning project for recognizing hand-written digits of 0-9 as their numeric values. The data comes from the [MNIST data-set] (http://yann.lecun.com/exdb/mnist), as part of a [Kaggle](https://www.kaggle.com/c/digit-recognizer) competition.
The code is developed in C# .NET, using the CsvHelper library for parsing the csv file of images, and Accord .NET for the gaussian SVM. The images are provided as 28x28 gray-scale pixel values (0-255), resulting in 784 pixels per image (ie., 784 columns in each csv row). An additional column is included for the label identifier.
For example (for the digit '7'):
7,0,0,0,0,255,127,0,0...
## Author
Kory Becker
http://www.primaryobjects.com/kory-becker.aspx