Handwritten Digit Recognition

Deep Learning model for Handwritten Digit Recognition using TensorFlow and Neural Network Techniques

1. Problem Definition ⛑

Recognise handwritten digit, from a dataset which contains B&W images of each digit written on 28x28 pixel box, given by The MNIST DATABASE

2. Dataset

The data we're using is officially provided by The MNIST DATABASE

The digits have been sized-normalized and centered in a fixed-sized image.

The data is quite preprocessed and well-formatted.

3. Evaluation

Accuracy should be above 95%.

4. Features

Some information about the data,

• We're dealing with images(unstructured data) so it's probably best we use deep learning/transfer learning technique to solve this problem.
• There are around a 60,000 examples of training set.
• There are around a 10,000 examples of test set.

5. Model/Estimator

Neural Network Multi-class Classification Model

6. Results

Last Successful Run: Accuracy was 96%

Idea (Improvisation) 💡

• Build a screen where kids can practice writing digits and machine will tell which number it was
• We can include formulas and signs in the dataset and train our model to convert it into a full fledged hand operated calculator where anyone can practice some mathematical formulas.