The augmented dataset must be formatted in a similar fashion to the MNIST dataset so that we can use the rust-autograd example (see mnist_data.rs and cnn_mnist.rs).
This might be complicated because our dataset resides in three layers of directories, so perhaps packing everything into a single object might be difficult. We may look to packing a single object per each language, or each language's character directory.
The augmented dataset must be formatted in a similar fashion to the MNIST dataset so that we can use the rust-autograd example (see
mnist_data.rs
andcnn_mnist.rs
).This might be complicated because our dataset resides in three layers of directories, so perhaps packing everything into a single object might be difficult. We may look to packing a single object per each language, or each language's character directory.