Digit Recognition using Softmax Regression

This code helps you classify different digits using softmax regression.

Code Requirements

You can install Conda for python which resolves all the dependencies for machine learning.

Description

Softmax Regression (synonyms: Multinomial Logistic, Maximum Entropy Classifier, or just Multi-class Logistic Regression) is a generalization of logistic regression that we can use for multi-class classification (under the assumption that the classes are mutually exclusive). In contrast, we use the (standard) Logistic Regression model in binary classification tasks.

For more information, see

Python Implementation

1. Dataset- MNIST dataset
2. Images of size 28 X 28
3. Classify digits from 0 to 9
4. Logistic Regression, Shallow Network and Deep Network Support added.

Train Acuracy ~ 91 to 99%

Test Acuracy ~ 90 to 98%

Execution for writing through webcam

To run the code, type python Dig-Rec.py

python Dig-Rec.py

Execution for showing images through webcam

To run the code, type python Digit-Recognizer.py

python Digit-Recognizer.py