Recognize Hand-Drawn Digital Digits

This model uses machine learning technology to recognize hand-drawn digital digits. The model has already been trained on 28*28 grayscale images as described in http://yann.lecun.com/exdb/mnist/. Hence, there is no need to train the model. You can start submitting images to model to be recognized.
Note Currently the model can only recognize (score) single digit images. The images must be 20 x 20 monochrome bitmaps. Also note that the model does not always correctly recognize the digits. This is because the model was trained on MNIST images. However, the model can be customized to recognize different image types if required. The model is based on code provided by Ruslan Salakhutdinov and Geoff Hinton. The original programs are available here.
Submitted By
Ruslan Salakhutdinov, Geoff Hinton
Cost
$0/month
Date Submitted
6/17/2013
Source Code
Available
Language
Python and Octave
Type
Classification
Algorithm
Deep auto-encoder artificial neural network
Customize this Model
Discuss this Model
Submit images for recognization

The easiest and quickiest way to test the model is to submit one of the following hand-drawn images

Submit your own hand-drawn image (20 x 20 Monochrome Bitmap only)
Submit the file below.
Upload File

Note:
You can create a hand-drawn image with tools such as Microsoft Paint