Deep learning/Keras Model over Semeion Handwritten Digit Image Data

Prezi:https://prezi.com/p/r1lkzdza1ffr/

Data Set Information:

1593 handwritten digits from around 80 persons were scanned, stretched in a rectangular box 16x16 in a gray scale of 256 values.Then each pixel of each image was scaled into a bolean (1/0) value using a fixed threshold. Each person wrote on a paper all the digits from 0 to 9, twice. The commitment was to write the digit the first time in the normal way (trying to write each digit accurately) and the second time in a fast way (with no accuracy).

Data Analysis:

• Implemented Keras Machine Learing algorithm to develop a deep net trained on the test data
• Trained the hyperparameters and performed model tuning by adjusting epoch and number of layers to optimize the outputs and make the performance better
• Developed the model with a performance accuracy of around 96% which classify the nuumeric image data to a resultant value

References:

The MNIST dataset is very well studied. Below are some additional resources you might like to look into.

• The Official MNIST dataset webpage
• Rodrigo Benenson’s webpage that lists state of the art results.
• Kaggle competition that uses this dataset (check the scripts and forum sections for sample code)
• Read-only model trained on MNIST that you can test in your browser (very cool)