Handwritten Digit Recognition Using TensorFlow Datasets - Data Every Day
Welcome back to Data Every Day!
Today's episode features the classic MNIST handwritten digit dataset. We will use TensorFlow Datasets to load up the data, and then process it with TensorFlow. Today's model is a simple neural network with 128 units in a single hidden layer.
Unfortunately, we ran into a problem with displaying the prediction results towards the end, but overall, we were successful! It's amazing that such a simple model can yield results of ~97-99% if the data is clean enough!
Here is a link to the TensorFlow Datasets MNIST data:
https://www.tensorflow.org/datasets/c...
And here is a link to the Google Colab notebook I used in the video:
https://colab.research.google.com/dri...
Thanks for watching! Remember to subscribe if you liked the video and stay tuned for more content! :)