Python Machine Learning Application
Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
README.md
test.py
test1.jpg
test2.JPG
test3.JPG
test_5.png
test_8.JPG
trainClassifier.py

README.md

Digit-Detector

Python Machine Learning Application

A commmand-line application developed to create a Nearest-Neighbors machine learning classifier that learns how to recognize single digits in images. The classifier is trained using the MNIST database.

  • trainClassifier.py downloads the database, trains the classifier, and saves it to the same directory.

  • test.py asks which images you'd like to analyze within the same directoy, displays the enhanced images that the classifier analyzes, and displays the classifier's digit predictions. The algorithm works best for clear images with a lot of contrast (e.g. black writing on white paper with no shadows).

Sample images are provided.

Currently, test.py performs fixed threshholding on the images. Future improvements include implimenting adaptive threshholding and recognizing multiple digits in one image.