Handwritten Digit Recognition
As part of F24 WRITING 101-06 DWoK assignment.
Draw a digit in the box above to see how the neural network recognizes digits.
The bars in the histogram reflects the relative values in the output layer, as discussed in the lesson.
The highest bar, or the neuron in the output layer with the largest numerical value, is the model's choice of digit it predicts.
Play around with the model to see if the model particularly struggles with any particular digits (think about what was previously discussed in the lesson). This model performs extremely well, however, as it is trained with more than 70k images. Below, you can find models that are trained on a much more limited set of training data, and are significant less accurate and confident in their predictions.
Model trained with only ~1500 images
Model trained with only ~300 pictures
Some numbers from the MNIST database that were used for training the neural network.
Picture by Josef Steppan. Used under a CC BY-SA 4.0 license.
Neural network model provided by © 2020 - IMAGINARY gGmbH. Used under a CC BY 4.0 license.