A TF.Keras Python Implementation

This is a simple implementation of the TableNet nerual net.

It is implemented as a python module that can be imported for training and usage.

Absolutely no guarantee that it works... or warrantee if it doesnt.

It is trained on a customized Marmot dataset.

This code has largely been based on the work of others (ex. https://github.com/jainammm/TableNet) but has been largely rewritten and expanded on to make it more generalized.

The initial marmot set used to train this NN is included here in a multi-part zip. It includes 509 images that have been anotated in the Marmot format. This includes column definition data.