Study on submodule structure of complex image modeling task.
We present a study on modeling the task of solving simple digit addition problems. The input to the model is an image consisting
of a simple addition problem such as 5 + 4, outputs an image containing the answer, in this case 9. This is a complex
model that we add apriori structure to, namely in the form of submodules. We investigate indpedent training of submodules
along with end to end training and evaluate the differences.
This repository contains the code for data generation, model training, saved models, and latex for the final paper report.
TODO
TODO