Prediction:
Digit Classification
A beginning.
Overview
As my first project, I wanted to create a digit classifier using the CNN architecture. My model used two convolution layers with feature layers 8 and 64 respectively, and a kernel size of 3. I also used pooling (for efficiency). There is a final single linear layer. I used Adam optimization, and trained for 4 epochs on the MNIST dataset.
Demo
Draw a digit on the black canvas.